AI engineer and data scientist in Bonn, Germany. Deep-dive guides that rebuild the AI stack from first principles.https://shukurza.de/https://shukurza.de/guides/transformers/the-machinery-how-a-neural-network-computes-and-learns/https://shukurza.de/guides/transformers/the-machinery-how-a-neural-network-computes-and-learns/A neural network is a function with adjustable numbers in it, and learning means adjusting those numbers to make a loss smaller. Everything in this guide, attention included, is built from three…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/making-deep-networks-trainable/https://shukurza.de/guides/transformers/making-deep-networks-trainable/Depth is powerful but fragile: a signal passing through 100 layers gets multiplied 100 times, so it tends to explode or vanish, and gradients do the same on the way back. Modern networks train…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/turning-text-into-numbers/https://shukurza.de/guides/transformers/turning-text-into-numbers/Neural networks operate on vectors; text is a string of symbols. Two steps bridge the gap: tokenization splits text into units and maps each to an integer id, and embedding maps each id to a learned…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/attention-from-first-principles/https://shukurza.de/guides/transformers/attention-from-first-principles/Attention lets each token build a new vector as a weighted average of information from other tokens, with weights computed from content. It is a soft, differentiable dictionary lookup, and it is the…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/the-transformer-block/https://shukurza.de/guides/transformers/the-transformer-block/A transformer is a stack of identical blocks. Each block does two things in turn: multi-head attention moves information between positions, then an MLP processes each position on its own. Both read…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/position-telling-the-model-where-each-token-is/https://shukurza.de/guides/transformers/position-telling-the-model-where-each-token-is/Attention treats its input as an unordered set, so word order must be injected explicitly. The method chosen decides how well a model handles sequences longer than those it trained on. Today the…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/architecture-families-and-training-objectives/https://shukurza.de/guides/transformers/architecture-families-and-training-objectives/The same block can be wired three ways, differing in which tokens may attend to which and what the model is trained to predict. Encoder-only models read bidirectionally and fill in masked tokens…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/pretraining-at-scale/https://shukurza.de/guides/transformers/pretraining-at-scale/Pretraining teaches a decoder to predict the next token over trillions of tokens of text and code, and almost all of a model's knowledge and raw capability comes from this stage. Three tools make it…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/inference-and-serving/https://shukurza.de/guides/transformers/inference-and-serving/Generating text runs the model once per output token, and that loop has two very different phases. Prefill processes the whole prompt in parallel and is limited by compute; decode produces one token…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/efficiency-variants-moe-long-context-and-beyond-attention/https://shukurza.de/guides/transformers/efficiency-variants-moe-long-context-and-beyond-attention/Two costs shape architecture research: dense models spend every parameter on every token, and attention's cost grows with the square of context length. Mixture-of-experts attacks the first by…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/post-training-from-text-predictor-to-assistant/https://shukurza.de/guides/transformers/post-training-from-text-predictor-to-assistant/A pretrained base model continues documents; it doesn't answer, follow instructions or decline anything. Post-training turns it into an assistant in stages: supervised fine-tuning on example…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/adapting-models-prompting-fine-tuning-lora-and-distillation/https://shukurza.de/guides/transformers/adapting-models-prompting-fine-tuning-lora-and-distillation/To make a model good at your task, change its inputs before you change its weights. Prompting and retrieval are cheap, fast to iterate and reversible; fine-tuning is for behaviour prompting can't…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/embeddings-retrieval-and-rag/https://shukurza.de/guides/transformers/embeddings-retrieval-and-rag/Retrieval-augmented generation gives a model the right information at query time: find relevant passages, put them in the prompt, and have the model answer from them. It is the standard way to use…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/beyond-text-vision-speech-and-multimodal-models/https://shukurza.de/guides/transformers/beyond-text-vision-speech-and-multimodal-models/A transformer processes a sequence of vectors and doesn't care where they came from. Cut an image into patches or audio into frames, project each piece to a vector, and the same architecture applies.…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/inside-the-model-interpretability/https://shukurza.de/guides/transformers/inside-the-model-interpretability/We can read every number in a transformer, yet understanding what it computes is an open research field. A few ideas have become standard vocabulary: the residual stream as a shared channel…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/evaluation-failure-modes-and-security/https://shukurza.de/guides/transformers/evaluation-failure-modes-and-security/Benchmarks tell you roughly how capable a model is; only an evaluation built from your own task tells you whether your system works. And because an LLM receives instructions and data through the same…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/from-model-to-system/https://shukurza.de/guides/transformers/from-model-to-system/An LLM call is a stateless function: a context goes in, a next-token distribution comes out, repeated until it stops. Memory, tools, agents and "knowledge" are all built by the surrounding system…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/transformers/fluency-kit/https://shukurza.de/guides/transformers/fluency-kit/Fluency means you can explain each idea from first principles, estimate its costs on the back of an envelope, and ask the question that exposes a weak claim. This section collects the vocabulary, the…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/the-model-how-german-is-built/https://shukurza.de/guides/german/the-model-how-german-is-built/German marks *who does what to whom* on the small words around a noun, not on word order. Almost every rule that feels hard (cases, article changes, adjective endings, free word order) follows from…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/sounds-and-spelling/https://shukurza.de/guides/german/sounds-and-spelling/German spelling is close to phonemic: with about fifteen rules you can pronounce any word you read and spell most words you hear. Vowel length is the rule that matters most, because it changes…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/nouns-gender-and-plural/https://shukurza.de/guides/german/nouns-gender-and-plural/Gender is a grammatical class, not biology, and most nouns can be predicted from their ending, their meaning or their last compound element. Learn the patterns below, then store every new noun as one…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/case-the-core-idea/https://shukurza.de/guides/german/case-the-core-idea/Case answers one question: what role does this noun play relative to the verb or the preposition that governs it? German marks four roles, and you already use all four in Azerbaijani; the difference…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/articles-and-declension-one-set-of-endings/https://shukurza.de/guides/german/articles-and-declension-one-set-of-endings/Every article form comes from one table of sixteen strong endings attached to a stem. Learn this table once and der, dieser, jeder, welcher, ein, kein, mein and the adjective endings all fall out of…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/prepositions-who-decides-the-case/https://shukurza.de/guides/german/prepositions-who-decides-the-case/After a preposition, the preposition alone decides the case. There are four groups: always Akkusativ, always Dativ, always Genitiv, and nine two-way prepositions that switch between Akkusativ…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/adjective-endings-signal-once/https://shukurza.de/guides/german/adjective-endings-signal-once/A noun phrase must show its gender and case once, with a strong ending from section 5. If the word before the adjective already shows it, the adjective takes a plain weak ending (-e or -en); if…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/pronouns-the-same-signals-new-stems/https://shukurza.de/guides/german/pronouns-the-same-signals-new-stems/Pronouns reuse the article signals, so they are not separate vocabulary: *er → ihn → ihm* is *der → den → dem*, and *sie → ihr* (feminine Dativ) is *die → der*. Learn the pattern and the forms follow.Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/verbs-stems-prefixes-and-frames/https://shukurza.de/guides/german/verbs-stems-prefixes-and-frames/Every verb form is a stem plus an ending, so you need only four facts per verb: the infinitive, the Präteritum stem, the Partizip II, and whether the Perfekt uses *haben* or *sein*. Add a fifth, the…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/word-order-the-sentence-frame/https://shukurza.de/guides/german/word-order-the-sentence-frame/Every German clause is built on one frame of five fields. The verb parts sit in two fixed brackets; everything else is arranged so that known information comes early and new or important information…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/tenses-and-how-they-are-really-used/https://shukurza.de/guides/german/tenses-and-how-they-are-really-used/German has six tenses on paper but a simpler system in practice: no progressive aspect, the present tense for most futures, and a choice between Perfekt and Präteritum that depends on register more…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/mood-necessity-unreality-reported-speech/https://shukurza.de/guides/german/mood-necessity-unreality-reported-speech/Mood marks how real or how certain a statement is. Modal verbs express ability, permission and obligation; Konjunktiv II marks the unreal, the hypothetical and the polite; Konjunktiv I marks reported…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/passive-and-its-alternatives/https://shukurza.de/guides/german/passive-and-its-alternatives/The passive moves attention from the doer to the event or the thing affected. Formal German uses it heavily, so C1 requires the process passive, the state passive, and the half-dozen constructions…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/word-formation-multiplying-your-vocabulary/https://shukurza.de/guides/german/word-formation-multiplying-your-vocabulary/German builds most of its vocabulary from a limited set of roots combined by transparent rules. Learning a root with its prefixes and suffixes gives you a whole family of words at once, and lets you…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/connectors-and-particles-structure-and-tone/https://shukurza.de/guides/german/connectors-and-particles-structure-and-tone/Two groups of small words separate B1 from C1: connectors that build an argument, and modal particles that carry attitude. Connectors make your writing structured; particles make your speech sound…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/register-which-german-where/https://shukurza.de/guides/german/register-which-german-where/Fluency at C1 means switching deliberately between registers: the same speaker uses different tenses, cases and vocabulary in a bar, a team meeting, a report and a letter to an authority.Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/how-acquisition-works-and-your-path-to-c1/https://shukurza.de/guides/german/how-acquisition-works-and-your-path-to-c1/Rules make you notice a structure; only large amounts of input and corrected output make it automatic. This guide supplies the noticing; the daily system below supplies the volume.Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/asking-the-right-questions/https://shukurza.de/guides/german/asking-the-right-questions/Nearly every German error traces back to five questions. Ask them in this order about any sentence you read or write, and you will either find the answer yourself or know exactly what to look up.Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/german/quick-reference-tables/https://shukurza.de/guides/german/quick-reference-tables/The two lookups you will use most: every adjective ending in one grid, and the forty most common strong and mixed verbs. The strong endings, article and pronoun tables are in sections 5 and 8, the…Sun, 01 Nov 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/the-core-mental-model/https://shukurza.de/guides/cloud/the-core-mental-model/The cloud is a giant shared computer that you control through an API. It exists because of one economic fact: most computers sit idle most of the time, and pooling them across thousands of customers…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/the-physical-layer/https://shukurza.de/guides/cloud/the-physical-layer/Every cloud resource is a real machine in a real building in a specific place. That place decides your latency, which country's law applies, what can fail together, and often the price.Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/virtualization-how-one-machine-becomes-many/https://shukurza.de/guides/cloud/virtualization-how-one-machine-becomes-many/Virtualization is the trick that turns one physical server into many isolated, rentable pieces. Virtual machines, containers and serverless are three points on one trade-off: stronger isolation and…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/compute-for-ai-gpus-and-accelerators/https://shukurza.de/guides/cloud/compute-for-ai-gpus-and-accelerators/A GPU is a machine for doing the same arithmetic on huge arrays at once. For AI, the number that usually limits you is not how fast it computes but how fast it can move bytes out of its own memory.…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/back-of-envelope-math-for-ai-workloads/https://shukurza.de/guides/cloud/back-of-envelope-math-for-ai-workloads/Six short formulas let you estimate memory, time and cost for almost any AI workload in a few minutes, before you rent anything. This is the skill that most visibly separates fluent engineers from…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/storage-and-data/https://shukurza.de/guides/cloud/storage-and-data/Cloud storage comes in three shapes: block, file and object. AI systems use all three, plus fast local disks. Object storage is the durable home for datasets, weights and checkpoints; local NVMe is…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/networking/https://shukurza.de/guides/cloud/networking/A cloud network is a private network you define in software, called a VPC. You carve it into subnets and connect it to the internet only through doors you choose to open. Most security incidents…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/identity-security-and-compliance/https://shukurza.de/guides/cloud/identity-security-and-compliance/Everything in the cloud is an API call, and every call is checked against one question: is this identity allowed to do this action on this resource? Get identity right and most of security follows…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/orchestration-running-fleets-not-machines/https://shukurza.de/guides/cloud/orchestration-running-fleets-not-machines/Orchestration is software that keeps a fleet of machines doing what you declared: placing work, restarting what fails, scaling up and down. Kubernetes dominates services and inference, Slurm…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/infrastructure-as-code-and-delivery/https://shukurza.de/guides/cloud/infrastructure-as-code-and-delivery/Infrastructure as code (IaC) means your cloud setup is written in text files that are versioned and reviewed like software. It can then be recreated exactly, audited, and torn down completely. Use…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/distributed-training-in-the-cloud/https://shukurza.de/guides/cloud/distributed-training-in-the-cloud/Distributed training splits one training job across many GPUs. The design question is always which tensors to split and over which link they travel, because communication, not computation, sets the…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/inference-and-serving/https://shukurza.de/guides/cloud/inference-and-serving/Serving is where most AI systems spend most of their money over their lifetime. It is a queueing and memory-management problem, built on the physics from Parts 4 and 5. The first decision is whether…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/reliability-and-observability/https://shukurza.de/guides/cloud/reliability-and-observability/Reliability is designing for failures you cannot prevent; observability is being able to explain what your system is doing from the outside. AI systems need both, plus three extra signals: GPU…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/cloud-economics/https://shukurza.de/guides/cloud/cloud-economics/Cloud cost is price times quantity, divided by how much of it you actually use. In AI systems the GPU and token lines dominate, and the cheapest GPU-hour is the one you did not leave running.Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/the-provider-landscape/https://shukurza.de/guides/cloud/the-provider-landscape/The market splits into four groups: hyperscalers with hundreds of services, GPU-focused clouds, serverless and inference platforms, and European providers. The concepts are identical across all of…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/fluency-toolkit-the-questions-to-ask/https://shukurza.de/guides/cloud/fluency-toolkit-the-questions-to-ask/Fluency is knowing which question to ask next. These checklists cover the situations you will meet most often; the last section says where trustworthy answers live. For a single unfamiliar service…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/hands-on-learning-path/https://shukurza.de/guides/cloud/hands-on-learning-path/Reading builds vocabulary; building builds fluency. These eight stages take roughly three to four months part-time and around €200–500 in cloud spend, if you tear everything down after each stage.…Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/glossary/https://shukurza.de/guides/cloud/glossary/The terms you will hear most, alphabetically, each with the part that explains it.Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cloud/resources-worth-your-time/https://shukurza.de/guides/cloud/resources-worth-your-time/A short list, chosen because each one teaches from first principles rather than product names. All are easy to find by title; most are free.Sat, 31 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/math-for-ai/the-language-of-mathematics/https://shukurza.de/guides/math-for-ai/the-language-of-mathematics/Most "I don't understand this paper" moments are really "I can't read this notation" moments. This pillar is the grammar every other pillar is written in.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/math-for-ai/linear-algebra/https://shukurza.de/guides/math-for-ai/linear-algebra/Linear algebra is the data structure of AI: every input, weight, activation and embedding is a vector, matrix or tensor. If you master one pillar deeply, make it this one.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/math-for-ai/calculus-and-matrix-calculus/https://shukurza.de/guides/math-for-ai/calculus-and-matrix-calculus/Calculus answers one question: if I nudge this, how much does that change? Training a model means turning billions of knobs, and calculus tells you which way to turn each one.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/math-for-ai/probability/https://shukurza.de/guides/math-for-ai/probability/A language model never outputs a word; it outputs a probability distribution over every possible next token. Probability is the mathematics of what a model actually produces.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/math-for-ai/statistics-and-evaluation/https://shukurza.de/guides/math-for-ai/statistics-and-evaluation/Probability goes from a known distribution to data. Statistics runs the other way: from finite data back to conclusions about the world. Training a model is estimation, and evaluating one is…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/math-for-ai/information-theory/https://shukurza.de/guides/math-for-ai/information-theory/Information theory measures surprise. It explains why cross-entropy is the loss for every language model, what perplexity means, and why prediction and compression are the same problem.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/math-for-ai/optimization/https://shukurza.de/guides/math-for-ai/optimization/Calculus tells you which way is downhill. Optimization is the craft of actually getting down a mountain with billions of dimensions, in fog, using noisy directions, on a limited budget.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/math-for-ai/numerical-and-systems-math/https://shukurza.de/guides/math-for-ai/numerical-and-systems-math/The first six pillars assume perfect numbers and free computers. This pillar is what happens on real hardware, with finite precision, finite memory and a budget. For AI systems work, it is where most…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/math-for-ai/capstone-one-transformer-training-step-through-every-pillar/https://shukurza.de/guides/math-for-ai/capstone-one-transformer-training-step-through-every-pillar/The pillars are not separate subjects; a single training step uses all eight. Walk through this sequence until you can explain every line without notes. When you can, you are fluent.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/math-for-ai/learning-path-resources-and-question-bank/https://shukurza.de/guides/math-for-ai/learning-path-resources-and-question-bank/Follow the pillars in dependency order, build something in every phase, and finish by building a small GPT. Durations assume about 6–8 focused hours a week; adjust freely.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/why-version-control-exists/https://shukurza.de/guides/git/why-version-control-exists/Version control answers one question: what did this project look like at any moment, and who changed what, when, and why? Without it you get train_final_v3_REALLY_final.py, lost work, and models…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/git-s-data-model-a-content-addressed-object-database/https://shukurza.de/guides/git/git-s-data-model-a-content-addressed-object-database/Inside .git/objects, Git stores exactly four kinds of objects, each named by a hash of its own contents. Branches, merges and history are all built from these four.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/references-branches-head-and-tags-are-pointers/https://shukurza.de/guides/git/references-branches-head-and-tags-are-pointers/A branch is a name that holds one commit hash, and nothing more. Once you see this, branching, merging and recovery stop being mysterious.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/the-three-areas-working-tree-index-repository/https://shukurza.de/guides/git/the-three-areas-working-tree-index-repository/Every change you make lives in one of three places, and every "where did my change go?" question is answered by asking which one.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/daily-commands-mapped-onto-the-model/https://shukurza.de/guides/git/daily-commands-mapped-onto-the-model/About fifteen commands cover nearly all daily work, and each one is just a read or a write on the objects, refs and areas from Parts 2–4.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/branching-and-integration-merge-rebase-cherry-pick/https://shukurza.de/guides/git/branching-and-integration-merge-rebase-cherry-pick/Integration answers one question: two lines of history split from a common ancestor, so how do you get one line containing both? Git gives two answers. Merge records the join; rebase replays one line…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/undoing-things-safely/https://shukurza.de/guides/git/undoing-things-safely/Every undo in Git is one of three moves: add a new commit that cancels an old one (revert), move a branch pointer (reset), or copy a stored version over your files (restore). Pick by asking two…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/remotes-and-collaboration/https://shukurza.de/guides/git/remotes-and-collaboration/A remote is simply another copy of the database, with a name and a URL. Collaboration is copying objects between databases and then asking each side to move its refs.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/workflows-and-conventions/https://shukurza.de/guides/git/workflows-and-conventions/A workflow is an agreement about which branches exist, who writes to them, and how a change reaches main. Git enforces none of it; hosts enforce some of it through branch protection. For solo and…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/git-for-ai-and-ml-work/https://shukurza.de/guides/git/git-for-ai-and-ml-work/Git is excellent for small text that changes line by line - code, configs, prompts - and poor at large binaries such as datasets and checkpoints. AI work therefore pairs Git with tools that keep big…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/git-with-ai-coding-agents/https://shukurza.de/guides/git/git-with-ai-coding-agents/When an agent writes the code, Git becomes your control system. Commits are checkpoints you can return to, diffs are where you review, branches and worktrees are sandboxes, and hooks and CI are…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/power-tools/https://shukurza.de/guides/git/power-tools/Most of Git's power tools answer one of two questions: "when, where and why did this change?" or "how do I work with a huge or composite repository?" Learn them once and they save days.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/under-the-hood-build-a-commit-by-hand/https://shukurza.de/guides/git/under-the-hood-build-a-commit-by-hand/Git's everyday commands, called porcelain, are built from small low-level commands called plumbing. Building a commit by hand with plumbing is the fastest way to make Parts 2–4 permanent in your head.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/debugging-mindset-and-recovery-playbook/https://shukurza.de/guides/git/debugging-mindset-and-recovery-playbook/When Git confuses you, stop typing commands and answer four questions. Where is HEAD? Where do the relevant refs point? What is in the index and the working tree? Is an operation in progress? Almost…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/practice-path-fluency-checklist-and-glossary/https://shukurza.de/guides/git/practice-path-fluency-checklist-and-glossary/Reading builds the model; drills make it fluent. Work through these exercises in order in ~/git-lab, and tick them off as you go.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/git/sources/https://shukurza.de/guides/git/sources/Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/why-ai-runs-on-data-engineering/https://shukurza.de/guides/data-engineering/why-ai-runs-on-data-engineering/A trained model is a compressed function of its training data, so a model can never be better than the data pipeline that fed it. Every AI system is a data system first and a model second.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/representing-data-bytes-types-schemas-formats/https://shukurza.de/guides/data-engineering/representing-data-bytes-types-schemas-formats/All data is bytes plus an agreement about how to read them. Choosing a representation is choosing which operations will be cheap and which will be expensive.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/storage-where-bytes-live-and-what-reading-them-costs/https://shukurza.de/guides/data-engineering/storage-where-bytes-live-and-what-reading-them-costs/Each step down the memory hierarchy is roughly 10 to 1,000 times slower and cheaper than the one above. Almost every storage design rule follows from that: read as little as possible from slow tiers…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/databases-from-first-principles/https://shukurza.de/guides/data-engineering/databases-from-first-principles/A database is a storage engine, a query engine and a set of guarantees bundled together. Knowing which guarantees it gives and which access pattern it optimizes tells you when to use it and when not…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/data-modeling-shaping-data-for-its-readers/https://shukurza.de/guides/data-engineering/data-modeling-shaping-data-for-its-readers/A data model decides which questions are easy and which are painful. You model for how the data will be read, and you choose deliberately between storing a fact once (and joining) or copying it (and…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/distributed-systems-the-physics-of-many-machines/https://shukurza.de/guides/data-engineering/distributed-systems-the-physics-of-many-machines/Distribute only when one machine cannot hold the data, cannot process it fast enough, or must not be a single point of failure. In exchange you inherit partial failure, network delay, and the absence…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/batch-processing-transforming-data-in-bulk/https://shukurza.de/guides/data-engineering/batch-processing-transforming-data-in-bulk/Batch processing runs a computation over a bounded dataset in one go. Its core moves, split the data, process the pieces in parallel, regroup by key, combine, have not changed since MapReduce in…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/streaming-data-that-never-ends/https://shukurza.de/guides/data-engineering/streaming-data-that-never-ends/Streaming processes unbounded data continuously, and it rests on one abstraction: the append-only log. Its hard problems are all about time: events arriving late, out of order, and needing memory…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/architectures-warehouse-lake-lakehouse/https://shukurza.de/guides/data-engineering/architectures-warehouse-lake-lakehouse/The three big architectures differ on one question: where structure and guarantees are enforced. The lakehouse became the default by adding database-style transactions on top of cheap, open files in…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/pipelines-and-orchestration/https://shukurza.de/guides/data-engineering/pipelines-and-orchestration/A pipeline is a graph of idempotent steps over partitioned data. Orchestration is the system that runs that graph in dependency order, on a schedule or on events, retries what fails, and lets you…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/data-quality-contracts-lineage-and-observability/https://shukurza.de/guides/data-engineering/data-quality-contracts-lineage-and-observability/Data fails silently: a broken pipeline usually runs green and produces plausible-looking wrong output. Trust therefore has to be engineered: define what correct means, test it at every boundary, make…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/data-engineering-for-machine-learning/https://shukurza.de/guides/data-engineering/data-engineering-for-machine-learning/ML data engineering protects one invariant: at training time the model must see exactly the information it will have at prediction time, computed exactly the same way. The second duty is being able…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/data-engineering-for-llms/https://shukurza.de/guides/data-engineering/data-engineering-for-llms/LLM training is the largest data engineering problem in AI. Pretraining turns petabytes of messy web pages, code and documents into a clean, deduplicated, filtered, mixed and tokenized stream…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/data-engineering-for-rag-and-ai-applications/https://shukurza.de/guides/data-engineering/data-engineering-for-rag-and-ai-applications/A retrieval-augmented generation (RAG) system is a search engine whose reader is a language model. Its answer quality is decided mainly by the data pipeline behind it: parsing, chunking, embedding…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/feeding-gpus-data-loading-for-training/https://shukurza.de/guides/data-engineering/feeding-gpus-data-loading-for-training/Accelerators are the most expensive resource in an AI system, and training runs only as fast as its slowest stage. The data pipeline's job is to deliver every batch before the GPU asks for it; an…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/governance-privacy-and-security/https://shukurza.de/guides/data-engineering/governance-privacy-and-security/Governance answers who may do what with which data, for what purpose, and for how long. It weighs more in AI than in reporting, because models can memorize and reproduce their training data, and…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/cost-and-performance-reasoning/https://shukurza.de/guides/data-engineering/cost-and-performance-reasoning/Data system cost is dominated by four things: bytes scanned, bytes stored, bytes moved and compute hours. Performance work and cost work are therefore mostly the same work, and both start with…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/the-tool-landscape-mapped-to-principles/https://shukurza.de/guides/data-engineering/the-tool-landscape-mapped-to-principles/Every tool answers one or two of the six questions from section 0, and makes a specific trade-off. Place a new tool on this map first; its details then become easy to learn. Names below are…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/learning-path-and-hands-on-projects/https://shukurza.de/guides/data-engineering/learning-path-and-hands-on-projects/Fluency comes from building small systems, breaking them on purpose, and fixing them. The path below moves from foundations to AI-specific scale; each phase ends with a gate you can test yourself…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/question-bank-for-fluency/https://shukurza.de/guides/data-engineering/question-bank-for-fluency/Fluency means asking the right question at the right moment. Use the first two lists in design reviews and incidents; use the self-test to check your understanding, covering the answer column first.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/data-engineering/glossary/https://shukurza.de/guides/data-engineering/glossary/Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/what-security-actually-is/https://shukurza.de/guides/cybersecurity/what-security-actually-is/A system is secure when it keeps doing what it should, and nothing else, even while an intelligent adversary is actively trying to make it misbehave. The adversary is the whole difference between…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/how-computers-fail/https://shukurza.de/guides/cybersecurity/how-computers-fail/To a computer, instructions and data are both just bytes; the difference exists only by convention. Most low-level attacks happen when an attacker makes the system cross that convention - data gets…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/networks/https://shukurza.de/guides/cybersecurity/networks/A network is a chain of machines you do not control, passing your bytes along. By default, any of them can read, drop, replay or modify what passes through. Network security is the art of not…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/cryptography/https://shukurza.de/guides/cybersecurity/cryptography/Cryptography turns the problem of protecting a lot of data into the problem of protecting a small key. It can provide confidentiality, integrity and authenticity, but never more security than the key…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/identity-and-access/https://shukurza.de/guides/cybersecurity/identity-and-access/Every request raises two separate questions: who are you (authentication), and are you allowed to do this (authorization)? Most serious breaches are failures of the second question, and AI agents…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/web-and-application-security/https://shukurza.de/guides/cybersecurity/web-and-application-security/Nearly every web vulnerability is either an injection (data treated as code) or a missing authorization check. The browser runs code from many sources side by side, and the server receives requests…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/cloud-infrastructure-supply-chain-detection/https://shukurza.de/guides/cybersecurity/cloud-infrastructure-supply-chain-detection/In the cloud, identity is the perimeter: whoever holds the right credentials controls the infrastructure. Most cloud breaches come from over-broad permissions, public resources and stolen keys, not…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/the-bridge-why-ai-systems-are-different/https://shukurza.de/guides/cybersecurity/the-bridge-why-ai-systems-are-different/AI systems inherit every classical security problem from Parts 1–7, then add a few genuinely new ones. The deepest change: behaviour is learned from data rather than written in code, so whoever…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/adversarial-machine-learning/https://shukurza.de/guides/cybersecurity/adversarial-machine-learning/Models can be fooled at inference time (evasion), corrupted during training (poisoning and backdoors), and made to leak their training data or themselves (privacy and extraction attacks). A decade of…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/llm-and-generative-ai-security/https://shukurza.de/guides/cybersecurity/llm-and-generative-ai-security/An LLM reads instructions and data as one stream of tokens, so any text it reads can act as an instruction. Prompt injection follows directly, and it still has no complete fix. Secure LLM…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/agentic-systems/https://shukurza.de/guides/cybersecurity/agentic-systems/An agent is an LLM in a loop with tools, memory and the authority to act. That turns prompt injection from "bad text" into "bad actions": an agent is a confused deputy that reads untrusted content…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/the-ml-supply-chain-and-mlops/https://shukurza.de/guides/cybersecurity/the-ml-supply-chain-and-mlops/Models and datasets are dependencies like any library, except they are harder to inspect, sometimes execute code when loaded, and are usually pulled from public hubs with weak provenance. Securing…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/privacy-governance-and-regulation/https://shukurza.de/guides/cybersecurity/privacy-governance-and-regulation/Regulation turns good security and privacy practice into obligations with deadlines and fines. For AI work in Germany, four regimes matter most: GDPR (any personal data), the EU AI Act (AI systems…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/red-teaming-evaluation-and-ai-for-defense/https://shukurza.de/guides/cybersecurity/red-teaming-evaluation-and-ai-for-defense/You cannot prove an AI system secure, so you measure it. Red teaming finds failures nobody imagined; evaluation tells you how often they happen and whether fixes hold. Both only count if they model…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/worked-example-threat-modeling-an-ai-agent/https://shukurza.de/guides/cybersecurity/worked-example-threat-modeling-an-ai-agent/This part applies everything above to one realistic system, using Shostack's four questions from Part 1. The verdict comes first: the biggest risk is an outsider's email steering the assistant into…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/the-question-toolkit/https://shukurza.de/guides/cybersecurity/the-question-toolkit/Fluency in security is mostly the ability to ask the right question at the right moment, and to tell a strong answer from a weak one. This part collects the universal questions, shows how to sharpen…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/learning-path-practice-reading-list/https://shukurza.de/guides/cybersecurity/learning-path-practice-reading-list/Reading builds vocabulary; building and breaking your own systems builds understanding. This part gives a 12-week path at roughly 6–8 hours per week, the hands-on resources behind it, a reading list…Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/glossary/https://shukurza.de/guides/cybersecurity/glossary/One-line definitions of the terms used throughout the guide, alphabetical. The part where each is explained in depth is in brackets.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/cybersecurity/sources/https://shukurza.de/guides/cybersecurity/sources/Current framework and regulatory facts were checked on 30 October 2026 against these pages. Everything else reflects established knowledge and the papers named in Part 17.Fri, 30 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/what-python-actually-is/https://shukurza.de/guides/python/what-python-actually-is/Python is a language specification; CPython is the program that actually runs it. When people say "Python is slow" or "Python has a GIL", they mean CPython, the reference implementation written in C.…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/names-objects-and-references/https://shukurza.de/guides/python/names-objects-and-references/Assignment in Python never copies data; it binds a name to an object. This single rule explains most of Python's "surprising" behavior.Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/built-in-types-and-what-they-cost/https://shukurza.de/guides/python/built-in-types-and-what-they-cost/Each built-in type is a specific data structure with specific costs, and picking the right one is often a 100× decision. Learn what each one is underneath, not just its methods.Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/control-flow-and-expressions/https://shukurza.de/guides/python/control-flow-and-expressions/Python's control flow is small; the leverage is in its idioms. Truthiness, comprehensions, unpacking and "ask forgiveness" error handling make up most idiomatic code.Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/functions-in-depth/https://shukurza.de/guides/python/functions-in-depth/A function is an object created at runtime by def. It carries its compiled code, its default values and a reference to the scope where it was defined. Callbacks, closures and decorators all follow…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/iteration-and-generators/https://shukurza.de/guides/python/iteration-and-generators/Every for loop runs the same two-method protocol, and generators let you write that protocol as a function that pauses. This is how you stream datasets and LLM tokens without loading everything into…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/classes-and-the-data-model/https://shukurza.de/guides/python/classes-and-the-data-model/A class is an object that makes objects, and Python's syntax is a set of hooks that call special ("dunder") methods on them. len(x) calls x.len(), x[i] calls x.getitem(i), and model(x) calls…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/context-managers-and-exceptions/https://shukurza.de/guides/python/context-managers-and-exceptions/Exceptions are how Python reports failure, and context managers guarantee cleanup when failure happens. Together they decide whether a crashed job leaves behind open files, held locks or half-written…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/modules-packages-imports-and-environments/https://shukurza.de/guides/python/modules-packages-imports-and-environments/import executes a file once, caches the resulting module object, and binds a name to it. Most "works on my machine" problems in AI are environment problems: the wrong interpreter, the wrong…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/type-hints-and-pydantic/https://shukurza.de/guides/python/type-hints-and-pydantic/Type hints are ignored by the interpreter but read by tools and libraries, and Pydantic turns them into runtime validation. In LLM systems that combination becomes your contract with a model's output.Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/the-performance-model/https://shukurza.de/guides/python/the-performance-model/Python is slow per operation, yet AI workloads in Python are fast because the heavy math runs in compiled kernels. Performance work is mostly moving loops out of the interpreter and keeping the…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/concurrency-and-parallelism/https://shukurza.de/guides/python/concurrency-and-parallelism/First classify the work: I/O-bound work spends its time waiting (network, disk, LLM APIs) and needs concurrency; CPU-bound work spends it computing and needs parallelism. The right tool follows from…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/numpy-the-array-model-everything-else-copies/https://shukurza.de/guides/python/numpy-the-array-model-everything-else-copies/A NumPy array is one contiguous block of raw numbers plus metadata (dtype, shape, strides) describing how to read it. PyTorch tensors and JAX arrays use the same model, so learning it here pays off…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/data-and-serialization/https://shukurza.de/guides/python/data-and-serialization/Choose data formats by who reads them and whether the source can be trusted: text formats for interchange, columnar formats for analytics, safe tensor formats for weights. Never deserialize a pickle…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/pytorch-mental-model/https://shukurza.de/guides/python/pytorch-mental-model/A PyTorch tensor is a NumPy-style array that also knows its device and can record the operations that produced it. Training is three moves: run the forward computation, let autograd walk that record…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/ai-systems-engineering-in-python/https://shukurza.de/guides/python/ai-systems-engineering-in-python/An LLM application is an ordinary networked Python service whose slowest, least reliable and most expensive dependency is the model API. Engineer it like one: typed boundaries, async I/O, timeouts…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/tooling-debugging-and-testing/https://shukurza.de/guides/python/tooling-debugging-and-testing/Fluency is mostly fast feedback: run small pieces interactively, step through failures in a debugger, and lock behavior in with tests. Code written by AI assistants needs the last step most of all.Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/asking-the-right-questions/https://shukurza.de/guides/python/asking-the-right-questions/Fluency means knowing which question exposes the problem. Use this part as a checklist whenever you read, review or prompt for code, whether you or an AI assistant wrote it.Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/roadmap-and-build-to-learn-projects/https://shukurza.de/guides/python/roadmap-and-build-to-learn-projects/Learn by building small versions of the systems you will use; each project forces one part of this guide into your hands. Move to the next phase only when you pass its exit check, and skip phases…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/python/glossary-and-resources/https://shukurza.de/guides/python/glossary-and-resources/These are the terms to use when you search or ask for help; the right word usually finds the right answer.Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/the-problem-docker-solves/https://shukurza.de/guides/docker/the-problem-docker-solves/Your code never runs on "Python"; it runs on a precise stack of interpreter, libraries, system packages, kernel and drivers. Docker packages everything above the kernel so that stack travels with the…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/what-a-container-really-is/https://shukurza.de/guides/docker/what-a-container-really-is/There is no "container" object in the Linux kernel. A container is an ordinary process that the kernel shows a restricted view of the world (namespaces), caps in resources (cgroups), roots in a…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/images-layers-hashes-and-registries/https://shukurza.de/guides/docker/images-layers-hashes-and-registries/An image is an ordered stack of read-only filesystem layers plus a small JSON config, every piece addressed by its SHA-256 hash. A container adds one thin writable layer on top, and a union…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/docker-s-architecture/https://shukurza.de/guides/docker/docker-s-architecture/The docker command is a thin client. The real work happens in a chain: the CLI calls the Docker Engine API, dockerd asks containerd to run a container, containerd starts a small shim, the shim calls…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/container-lifecycle/https://shukurza.de/guides/docker/container-lifecycle/A container lives exactly as long as its main process, PID 1. Starting, stopping, restarting, logging and exit codes all follow from that one rule.Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/dockerfiles-in-depth/https://shukurza.de/guides/docker/dockerfiles-in-depth/A Dockerfile is a build program whose output is an image. Writing good ones comes down to three things: controlling the build context, ordering instructions for the layer cache, and keeping…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/storage-where-data-lives/https://shukurza.de/guides/docker/storage-where-data-lives/The container's own filesystem is disposable and slow for heavy writes. Anything large or worth keeping (datasets, model weights, checkpoints, caches, databases) goes in a volume or a bind mount…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/networking/https://shukurza.de/guides/docker/networking/Each container gets its own network namespace: its own interfaces, IP address, routing table and localhost. Docker connects those namespaces to each other and to the outside world with virtual…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/docker-compose-for-multi-service-ai-apps/https://shukurza.de/guides/docker/docker-compose-for-multi-service-ai-apps/Compose replaces a pile of docker run commands with one declarative YAML file, so an AI application made of several services (an API, a model server, a vector database, a cache, a background worker)…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/gpus-in-containers/https://shukurza.de/guides/docker/gpus-in-containers/To a container, an NVIDIA GPU is a set of device files (/dev/nvidia0, /dev/nvidiactl, /dev/nvidia-uvm) plus driver libraries that must exactly match the host's kernel driver. The NVIDIA Container…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/building-good-images-for-ai-workloads/https://shukurza.de/guides/docker/building-good-images-for-ai-workloads/A good AI image is pinned, built from a lockfile, small enough to pull quickly, built for the CPU architecture it will run on, and deliberate about where model weights live. Each of those is a…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/serving-and-production-patterns/https://shukurza.de/guides/docker/serving-and-production-patterns/In production a model container is a replaceable unit: configured by environment variables, honest about whether it's ready, bounded by resource limits, observable through logs and metrics, and able…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/security/https://shukurza.de/guides/docker/security/Containers isolate well enough for most workloads, but more weakly than virtual machines, because every container shares the host kernel. The real-world risks are rarely exotic escapes; they're…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/debugging-and-introspection/https://shukurza.de/guides/docker/debugging-and-introspection/Debug a container by asking four questions in order: did it start, what did it print, what does it see from inside, and which limit or permission stopped it. Each question has a command, and most…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/beyond-docker/https://shukurza.de/guides/docker/beyond-docker/Docker is the on-ramp. The OCI images you build flow unchanged into other runtimes and orchestrators, and university HPC clusters often don't allow Docker at all, so knowing the neighbours is part of…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/hands-on-learning-path/https://shukurza.de/guides/docker/hands-on-learning-path/Fourteen labs, each proving one idea from the guide with your own hands. Labs 1 to 10 run on your Mac; 11 and 12 need a rented Linux GPU machine; 13 and 14 run anywhere. Keep a short lab notebook…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/question-bank/https://shukurza.de/guides/docker/question-bank/Thirty questions, roughly in order of depth. Answer each out loud before reading the answer; if you can explain all thirty from the mechanisms rather than from memory, you're fluent. The answers are…Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/guides/docker/glossary-and-cheat-sheet/https://shukurza.de/guides/docker/glossary-and-cheat-sheet/Mon, 26 Oct 2026 00:00:00 GMThttps://shukurza.de/writing/stop-unit-testing-your-pipelines/https://shukurza.de/writing/stop-unit-testing-your-pipelines/Pipelines fail because your assumptions about the data are wrong, not because your code is. Test the assumptions.Mon, 12 May 2025 00:00:00 GMThttps://shukurza.de/writing/what-i-learned-rebuilding-attention-in-one-weekend/https://shukurza.de/writing/what-i-learned-rebuilding-attention-in-one-weekend/Mon, 28 Apr 2025 00:00:00 GMThttps://shukurza.de/writing/b1-in-six-months-german-grammar-that-compiles/https://shukurza.de/writing/b1-in-six-months-german-grammar-that-compiles/Sun, 30 Mar 2025 00:00:00 GMThttps://shukurza.de/writing/the-case-for-boring-orchestration/https://shukurza.de/writing/the-case-for-boring-orchestration/Mon, 17 Feb 2025 00:00:00 GMThttps://shukurza.de/writing/a-year-of-shipping-llms-to-production-in-numbers/https://shukurza.de/writing/a-year-of-shipping-llms-to-production-in-numbers/Sun, 08 Dec 2024 00:00:00 GMThttps://shukurza.de/writing/postgres-is-enough-until-it-isnt/https://shukurza.de/writing/postgres-is-enough-until-it-isnt/Mon, 21 Oct 2024 00:00:00 GMThttps://shukurza.de/writing/reading-papers-with-a-notebook-not-a-highlighter/https://shukurza.de/writing/reading-papers-with-a-notebook-not-a-highlighter/Sun, 04 Aug 2024 00:00:00 GMT