Posts
All the articles I've posted.
Designing AI Agents That Survive Production
A detailed architecture guide for production AI agents with typed tools, bounded autonomy, state machines, approvals, security, observability, and evaluation.
How Databases Actually Store Data: Pages, Indexes, and WAL
A detailed systems-level tour of database pages, row layouts, B-trees, buffer pools, MVCC, write-ahead logging, recovery, and performance tradeoffs.
Gemini: How Google Built a Multimodal AI Model
How Gemini-style multimodal models are trained and served: text, image, audio, video, long context, tool use, MoE ideas, and how open models like Kimi are designed.
Vector Search and RAG from First Principles
A detailed guide to embeddings, vector indexes, chunking, hybrid retrieval, reranking, citations, evaluation, and secure production RAG systems.