Learning Hub
Structured technical notes, deep dives, and references for modern backend and distributed systems.
All topics
15 topicsBackend
2 topics- Applied Systems DesignA repeatable method for any blank-page system design — interview or real review.
- REST APIWhat makes an API RESTful? Learn Roy Fielding's six REST constraints — statelessness, uniform interface, HATEOAS, cacheability — and how to nail the interview answer.
Data
3 topics- Bloom FiltersA Bloom filter trades certainty for space: "definitely not in the set" is guaranteed, "probably in the set" is not.
- Storage EnginesLSM-tree compaction in depth: leveled vs tiered strategies, the write/read/space amplification trade-off, and the tombstone resurrection bug — pick per table.
- Transactions & IsolationMVCC internals: how databases store row version chains, how a snapshot decides which version it sees, and why one long-running transaction causes table bloat and vacuum pressure.
Architecture
3 topics- Distributed SystemsHow distributed consensus works: the FLP impossibility result, Paxos, Raft leader election and safety, quorums, and the commit rules that keep replicated state machines correct.
- Production AI ArchitectureReview an AI feature the way you review any production service: reliability, scalability, observability — with one new variable, non-determinism, changing every lens.
- Streaming & Event-DrivenExactly-once semantics in Kafka explained: idempotent producers, transactions, and committing consumer offsets inside the producer transaction to make read-process-write loops exactly-once.
AI
5 topics- Advanced AI SystemsThe hype-free take on multi-agent systems: orchestrator-worker vs swarms, the coordination tax, when one well-tooled agent wins, and the anti-patterns.
- AI AgentsAI agents from first principles: how LLMs, tools, the agent loop, planning, and MCP fit together — build the full stack from memory and tell real capability from hype.
- AI EvaluationWhy evaluation is the real moat: the offline vs online split, golden sets, LLM-as-judge scoring, and the eval loop that turns failures into a regression suite.
- Context EngineeringExplain why an LLM cannot see your whole codebase, what actually fills the context window, and why context retrieval — not the model — is where modern AI systems win or lose.
- Domain Agent DesignDesign any business agent with one repeatable 9-part template — responsibilities, tools, context, memory, runtime, eval, security, approval gates, ROI.
Platform
1 topic- AI InfrastructureTrace one request through the AI infra stack: text to embeddings to vector DB (HNSW/IVF) to reranker to model gateway to inference server.
Security
1 topic- AI SecurityWhy AI security is different: the prompt is not a trust boundary.