EPHIZEN · A wiki for CS — your lifelong companion · Free & open
The CS knowledge map you can grep, search, and ship from.
Ephizen is a reference for working engineers — programming, systems, math, machine learning, and the modern AI stack — written in a single voice and cross-linked so you can land on any topic from search and leave with what you came for. No enrolment, no completion tracking, no "next chapter".
§ the roadmap
The CS Skill Tree
Twelve domains, from print("hello world") to shipping
agents in production. Click any leaf to jump to its guide; WIP nodes
are on the wiki’s roadmap.
12 domains 499 topics 239 linked 260 WIP
-
Programming
Pick a language, learn its idioms, master the tooling. Python is the wiki's default because the ML stack is Python, but the concepts transfer.
Pick a language- Python (wiki default) Guide
- Go WIP
- Rust WIP
- Java WIP
- JavaScript / TypeScript WIP
- C / C++ WIP
- C# WIP
Language fundamentals- Variables, types, control flow WIP
- Functions, scope, closures WIP
- Mutability, identity, references Guide
- List / dict / set / tuple Guide
- List & dict comprehensions Guide
- Error handling & exceptions WIP
- File I/O WIP
- Modules & imports WIP
OOP & idioms- Classes, inheritance, polymorphism WIP
- Composition over inheritance WIP
- Encapsulation & visibility WIP
- Dataclasses Guide
- Type hints Guide
- Context managers Guide
- Decorators Guide
- Iterators & generators Guide
- Logging Guide
Tooling & version control- venvs · pip · uv · poetry Guide
- Git fundamentals WIP
- Branching, rebasing, merge WIP
- GitHub / GitLab workflows WIP
- Editor / IDE setup WIP
- Pydantic v2 validation Guide
- pydantic Tool
Production patterns- pytest & fixtures Guide
- Profiling Guide
- async / await Guide
- asyncio for LLMs Guide
- Packaging & distribution WIP
- Code style & linting WIP
-
Data Structures
The containers algorithms run on. Know them cold — you'll choose them in every interview and every hot loop.
Linear- Array (fixed) & dynamic array WIP
- Linked list (reversal) Guide
- Doubly linked list WIP
- Stack Guide
- Queue & circular queue WIP
- Deque WIP
- Hash table / hashmap WIP
- Hash set WIP
- Collision strategies (chain / open addressing) WIP
Trees- Binary tree WIP
- BST WIP
- AVL tree (self-balancing) WIP
- Red-Black tree WIP
- B-tree (DB engine) WIP
- 2-3 / 2-3-4 / K-ary tree WIP
- Heap & priority queue (top-K) Guide
- Trie Guide
- Segment tree, Fenwick (BIT) WIP
- K-D tree WIP
Graphs- Adjacency list / matrix WIP
- Directed / undirected WIP
- Weighted graphs WIP
- DAGs WIP
- Spanning tree WIP
- Tree BFS Guide
- Tree DFS Guide
Specialised & probabilistic- Monotonic stack Guide
- Monotonic deque Guide
- Union-find (DSU) Guide
- Skip list WIP
- Bloom filter WIP
- Count-min sketch WIP
- HyperLogLog WIP
- LRU / LFU / MFU cache WIP
Asymptotic notation- Big O · Big Θ · Big Ω WIP
- Common runtimes (constant → factorial) WIP
- Amortised analysis WIP
- Space complexity WIP
-
Algorithms
Patterns that show up in every interview, every backend, every ML pipeline. Big-O is the language they speak.
Recursion & searching- Tail vs non-tail recursion WIP
- Binary search Guide
- Linear search WIP
- Graph BFS Guide
- Graph DFS Guide
- Bidirectional BFS WIP
Sorting- Bubble sort WIP
- Selection sort WIP
- Insertion sort WIP
- Merge sort WIP
- Quick sort WIP
- Heap sort WIP
- Counting / radix / bucket WIP
- Cyclic sort Guide
- K-way merge Guide
- Topological sort Guide
Coding patternsGraph algorithms- Dijkstra's shortest path Guide
- Bellman-Ford WIP
- Floyd-Warshall WIP
- A* search WIP
- Prim's MST WIP
- Kruskal's MST WIP
- Tarjan / Kosaraju (SCC) WIP
- Network flow (Ford-Fulkerson) WIP
Dynamic programming- 1D DP Guide
- Knapsack family Guide
- Grid & strings DP Guide
- Interval DP WIP
- Bitmask DP WIP
- Tree DP WIP
String search- Brute force search WIP
- Knuth-Morris-Pratt (KMP) WIP
- Rabin-Karp WIP
- Boyer-Moore WIP
- Suffix arrays WIP
- Aho-Corasick WIP
Greedy classics & backtracking puzzles- Huffman coding WIP
- Interval scheduling WIP
- N-Queens WIP
- Hamiltonian path WIP
- Knight's tour WIP
- Sudoku solver WIP
Complexity classes- P · NP · NP-complete · NP-hard WIP
- Reductions WIP
- Approximation algorithms WIP
- Randomised algorithms WIP
-
Computer Systems
What actually runs your code — CPU, memory, OS, concurrency. The layer most engineers wave at and most bugs come from.
How computers work- How a CPU executes programs WIP
- Registers & RAM WIP
- Instruction sets (x86, ARM, RISC-V) WIP
- CPU cache (L1/L2/L3) WIP
- How computers calculate (binary, two's complement) WIP
- Endianness (big / little) WIP
- Floating-point math (IEEE 754) WIP
- Character encodings (Unicode, UTF-8, ASCII) WIP
Operating systems- Processes vs threads WIP
- Process forking WIP
- Scheduling algorithms WIP
- CPU interrupts WIP
- Virtual memory & paging WIP
- Memory allocation (stack vs heap) WIP
- Syscalls WIP
- Filesystems WIP
- Linux fundamentals WIP
Concurrency- GIL · threading · multiprocessing Guide
- async / await Guide
- Coroutines & event loops WIP
- Producer / consumer WIP
- Locks, mutexes, semaphores WIP
- Deadlock & race conditions WIP
- Lock-free data structures WIP
- Concurrency in multiple cores WIP
-
Networking & Web
How packets become products. From wires to APIs to the LLM-serving stack you'll ship.
The internet- OSI 7-layer model WIP
- TCP/IP model WIP
- DNS WIP
- HTTP/1.1, /2, /3 WIP
- TLS / HTTPS WIP
- Cookies, sessions, CORS WIP
- Sockets WIP
Real-time data- WebSockets WIP
- Server-Sent Events (SSE) WIP
- Long polling vs short polling WIP
- WebRTC WIP
API styles- REST WIP
- GraphQL WIP
- gRPC WIP
- JSON-API · HATEOAS WIP
- OpenAPI / Swagger WIP
- postman Tool
Backend stack- FastAPI Tool
- Web servers (nginx, Caddy) WIP
- Reverse proxies & load balancers WIP
- CDNs WIP
- Auth: JWT, OAuth, sessions WIP
- Rate limiting, throttling WIP
- Backpressure & circuit breakers WIP
Message brokers & search- kafka Tool
- RabbitMQ WIP
- NATS WIP
- Elasticsearch WIP
- Solr WIP
System design- Horizontal vs vertical scaling WIP
- Caching (client / server / CDN) WIP
- CAP theorem · PACELC WIP
- Consensus (Raft, Paxos) WIP
- Sharding & partitioning strategies WIP
- Saga / 2PC (distributed txns) WIP
- Event sourcing · CQRS WIP
- Twelve-factor app WIP
Security basics- OWASP Top 10 WIP
- Hashing vs encryption WIP
- Symmetric vs asymmetric crypto WIP
- Password hashing (bcrypt, scrypt, argon2) WIP
- Secrets management WIP
-
Databases
Where state lives. Get fluent in SQL, learn the engine underneath, then pick the right shape for the job.
SQL fundamentalsAdvanced SQL- Window functions Guide
- Ranking (ROW_NUMBER, RANK, DENSE_RANK) Guide
- LAG / LEAD / FIRST_VALUE Guide
- Recursive CTEs Guide
- Time-series queries Guide
- JSON in SQL Guide
- Star schema & feature stores Guide
- Stored procedures & triggers WIP
- Views & materialised views WIP
Engine internals- Indexes (B-tree, hash, GIN) Guide
- EXPLAIN & query plans Guide
- Transactions & ACID Guide
- BASE & PACELC WIP
- Isolation levels Guide
- Locking (optimistic vs pessimistic) WIP
- Normalisation Guide
- Sharding & partitioning Guide
- OLTP vs OLAP Guide
- Replication (leader/follower, multi-master) WIP
- Database federation WIP
- Migrations WIP
Engine families- Relational (PostgreSQL, MySQL) WIP
- Document (mongodb, CouchDB) Tool
- Key-value (redis, DynamoDB) Tool
- Realtime (Firebase, RethinkDB) WIP
- Time-series (Influx, TimescaleDB) WIP
- Columnar (Cassandra, ScyllaDB) WIP
- Graph (Neo4j, Neptune) WIP
- Warehouses: snowflake Tool
- Lakehouse: databricks Tool
-
Math & Numerical Computing
The mathematical backbone of ML — plus the array libraries that let you ship it.
Discrete math- Logic & proof techniques WIP
- Set theory WIP
- Combinatorics WIP
- Graph theory WIP
- Number theory (modular arithmetic) WIP
- Information theory (entropy, MI) WIP
Linear algebra- Vectors, norms, dot products WIP
- Matrices & matrix multiplication WIP
- Determinants & inverses WIP
- Eigenvalues & eigenvectors WIP
- SVD & PCA derivation WIP
- Tensor operations WIP
Calculus- Derivatives & partial derivatives WIP
- Gradients & Jacobians WIP
- Chain rule (the ML one) WIP
- Hessians & second-order methods WIP
- Constrained optimisation (Lagrange) WIP
Probability & statistics- Random variables, distributions WIP
- Expectation, variance, covariance WIP
- CLT, law of large numbers WIP
- Bayes rule & conditional probability WIP
- MLE & MAP estimation WIP
- Hypothesis testing WIP
- A/B testing WIP
- CUPED & variance reduction WIP
- Bootstrap & resampling WIP
- KL divergence Term
-
Classical ML
The pre-deep-learning toolbox you still reach for first on tabular data. Half of "AI" in production is still gradient-boosted trees.
Supervised learning- Linear regression WIP
- Logistic regression WIP
- Naive Bayes WIP
- k-Nearest Neighbours WIP
- SVM (linear + kernel) WIP
- Decision trees WIP
- Random forests WIP
- Gradient boosting (XGBoost, LightGBM, CatBoost) Term
- xgboost Tool
Unsupervised- k-means Term
- Hierarchical clustering WIP
- DBSCAN WIP
- Gaussian mixtures WIP
- PCA WIP
- t-SNE / UMAP WIP
- Anomaly detection WIP
Time series & recsys- ARIMA / SARIMA WIP
- Prophet / NeuralProphet WIP
- Exponential smoothing WIP
- Collaborative filtering WIP
- Matrix factorisation WIP
- Two-tower retrieval WIP
Workflow & evaluation- Feature engineering Term
- Feature stores Term
- Cross-validation Term
- Bias / variance trade-off Term
- Overfitting Term
- Underfitting Term
- Regularisation (L1 / L2 / ElasticNet) Term
- Hyperparameter tuning Term
- Class imbalance & resampling WIP
- Calibration (Platt, isotonic) WIP
Metrics- Precision & recall Term
- F1 score Term
- ROC / AUC Term
- PR curve WIP
- Log loss / Brier score WIP
- MSE · MAE · RMSE · R² WIP
-
Deep Learning
Gradients all the way down. From the perceptron to ResNets and ViTs — the substrate every modern model is built on.
FoundationsOptimisers & regularisation- Optimiser (SGD, Momentum, RMSProp) Term
- Adam (paper) Paper
- AdamW · Lion · Sophia WIP
- Dropout Term
- Batch normalisation Term
- BatchNorm (paper) Paper
- LayerNorm · RMSNorm · GroupNorm WIP
- Weight decay & gradient clipping WIP
Architectures- MLP / perceptron WIP
- CNNs Term
- AlexNet (paper) Paper
- ResNet (paper) Paper
- U-Net (segmentation) WIP
- EfficientNet WIP
- RNN / LSTM / GRU WIP
- Vision Transformer Paper
- GAN Term
- Diffusion model Term
- Stable Diffusion (paper) Paper
- YOLO Term
Training techniques- Transfer learning Term
- Distillation Term
- Epoch / batch / step Term
- Mixed precision (FP16 / BF16) WIP
- Gradient accumulation WIP
- Data augmentation WIP
- Curriculum learning WIP
Frameworks -
Transformers & LLMs
Attention, tokenisation, pretraining, alignment — the engine room of every frontier model.
The transformerTokens & embeddingsTraining & alignmentFoundation models- Foundation model Term
- BERT (paper) Paper
- GPT-3 (paper) Paper
- LLaMA (paper) Paper
- CLIP (multimodal) Paper
- Anthropic Claude (Opus, Sonnet, Haiku) WIP
- OpenAI GPT-4 / GPT-4o / o1 WIP
- Google Gemini WIP
- Mistral & Mixtral WIP
- Cohere Command WIP
Inference & decodingServing & SDKs- OpenAI Chat Completions API WIP
- Anthropic Messages API WIP
- Tool use / function calling WIP
- Structured outputs (JSON mode) WIP
- Streaming responses WIP
- Token counting & pricing WIP
- vllm (high-throughput serving) Tool
- ollama (local LLMs) Tool
- huggingface-transformers Tool
Multimodal- Vision-language models (CLIP, LLaVA) WIP
- OpenAI Vision API (GPT-4V) WIP
- DALL-E / Midjourney / Imagen WIP
- Whisper (speech-to-text) WIP
- TTS (text-to-speech) WIP
- Audio understanding WIP
- Video understanding (Sora, Veo) WIP
-
RAG & Retrieval
Give the model your data. Embed, search, rerank, generate — the bread-and-butter pattern of every LLM app at work.
RAG fundamentals- RAG Term
- RAG (Lewis 2020 paper) Paper
- Chunking strategies Term
- Embedding Term
- Sentence Transformers WIP
- OpenAI embeddings API WIP
- Vector database Term
- HNSW (graph index) Term
- IVF / IVF-PQ WIP
- Similarity search (cosine, dot, L2) Term
Advanced retrieval- Hybrid search (dense + sparse / BM25) Term
- Reranker Term
- Reranking pipelines Term
- Graph RAG Term
- Agentic RAG Term
- Query rewriting / decomposition WIP
- MMR (max marginal relevance) WIP
- Citations & source attribution WIP
- DPR (paper) Paper
- ColBERT (paper) Paper
Vector databases- pinecone Tool
- chroma Tool
- weaviate Tool
- qdrant Tool
- lancedb Tool
- pgvector (Postgres extension) Tool
- Milvus WIP
- FAISS WIP
-
Agents, Eval & MLOps
Ship it. Prompts, agents, evaluation, monitoring, deployment — the production stack for LLM systems.
PromptingAgent patterns- Agent loop (perceive → reason → act) Term
- Multi-agent system Term
- Model Context Protocol (MCP) Term
- Tool use & function calling WIP
- Planner / executor split WIP
- Memory (episodic, semantic, scratchpad) WIP
Agent frameworksEvaluationSafety- Prompt injection Term
- Jailbreaks & adversarial prompts WIP
- Guardrails Term
- Hallucination Term
- Red teaming Term
- Bias & fairness audits WIP
- PII filtering & redaction WIP
- Content moderation APIs WIP
MLOps & deployment- LLMOps Term
- Model registry Term
- Canary / blue-green / shadow deploys Term
- Data drift Term
- Concept drift Term
- Drift detection (PSI, KS) Term
- bentoml (model serving) Tool
- ray-serve Tool
- docker Tool
- kubernetes Tool
- terraform (IaC) Tool
- GPU optimisation (Triton, TensorRT) WIP
- Workflow engines (Airflow, Prefect, Dagster) WIP
§ browse