/ writing

Notes from the build.

Engineering deep-dives drawn from real production work — AI agents, backend systems, payments, and the modern web. How things were actually built, trade‑offs included.

LatestJun 25, 2026 8 min

Model Context Protocol (MCP), explained for people who already build agents

Model Context Protocol explained for agent builders: what an MCP server is, tools vs resources vs prompts, a minimal TypeScript example, and MCP vs tool-calling.

AI AgentsMCPLLMsRead article

/ more posts

Jun 24, 2026 7 min
Metadata-filtered RAG: two-stage retrieval that stops returning irrelevant chunks
Metadata-filtered RAG fixes single-shot retrieval that returns junk on multi-topic corpora. How I built a metadata pre-filter, vector search, and LLM rerank pipeline.
RAGLLMsPythonRead
Jun 23, 2026 7 min
Hardening a Razorpay integration: signatures, idempotent webhooks, and a settlement ledger
Hardening a Razorpay integration in Next.js: checkout vs webhook signature verification, idempotent settlement with a Postgres ledger, and the operational guards.
PaymentsRazorpayNext.jsRead
Jun 22, 2026 8 min
The Next.js 15 caching model that finally clicked
Next.js 15 caching as one mental model: Request Memoization, Data Cache, Full Route Cache, Router Cache, and tag-based revalidation that actually fires.
Next.jsReactWeb PerformanceRead
Jun 21, 2026 7 min
Preventing double-bookings: race-free slot reservations in Postgres
Preventing double-bookings in Postgres: how a check-then-insert race oversells slots, and the layered fix — unique constraint, row and advisory locks, transactions, isolation.
PostgresSupabaseNext.jsRead
Jun 20, 2026 9 min
Vector search in Postgres with pgvector: embeddings, HNSW, and metadata filters
pgvector turns Postgres into a vector database: store embeddings, rank by cosine distance with <=>, index with HNSW, and filter by metadata in one SQL query.
PostgrespgvectorRAGRead
Jun 19, 2026 8 min
LLM structured outputs: how to force reliable JSON with schemas and tool calling
LLM structured outputs explained: force reliable JSON from OpenAI with JSON mode, tool calling, and json_schema strict decoding, then validate with Zod or Pydantic.
LLMsAIOpenAIRead
Jun 18, 2026 3 min
Making a portfolio AI-readable: llms.txt, structured data, and a crawler policy
Search is splitting into two audiences — Google's crawler and AI assistants like ChatGPT, Claude, and Perplexity. Here's exactly how I made this site legible to both, with a single source of truth so nothing drifts.
SEOAINext.jsRead
May 28, 2026 4 min
Building a human-in-the-loop news desk: a two-phase story-discovery agent
An editorial team's morning routine — search dozens of sources, judge what matters, confirm it, then draft — turned into a two-phase, human-in-the-loop agent streamed live over SSE. Here's the architecture and the decisions that kept it cheap and trustworthy.
AI AgentsPythonFastAPIRead

Subscribe via RSS

9 essaysGet in touch

Notes from the build.

Model Context Protocol (MCP), explained for people who already build agents

Metadata-filtered RAG: two-stage retrieval that stops returning irrelevant chunks

Hardening a Razorpay integration: signatures, idempotent webhooks, and a settlement ledger

The Next.js 15 caching model that finally clicked

Preventing double-bookings: race-free slot reservations in Postgres

Vector search in Postgres with pgvector: embeddings, HNSW, and metadata filters

LLM structured outputs: how to force reliable JSON with schemas and tool calling

Making a portfolio AI-readable: llms.txt, structured data, and a crawler policy

Building a human-in-the-loop news desk: a two-phase story-discovery agent

Notes from the build.

Model Context Protocol (MCP), explained for people who already build agents

Metadata-filtered RAG: two-stage retrieval that stops returning irrelevant chunks

Hardening a Razorpay integration: signatures, idempotent webhooks, and a settlement ledger

The Next.js 15 caching model that finally clicked

Preventing double-bookings: race-free slot reservations in Postgres

Vector search in Postgres with pgvector: embeddings, HNSW, and metadata filters

LLM structured outputs: how to force reliable JSON with schemas and tool calling

Making a portfolio AI-readable: llms.txt, structured data, and a crawler policy

Building a human-in-the-loop news desk: a two-phase story-discovery agent