Production systems that retrieve, reason, and respond. Built around real users, not benchmarks.
- LangChain
- LangGraph
- OpenAI
- Anthropic
- Azure AI
- RAG
- Vector DBs
- Pinecone
- pgvector
- Evals
I'm Aswin AK — an AI engineer & full-stack dev. I design and ship RAG pipelines, agentic workflows, and the interfaces around them. No magic. No vibes. Just systems that hold up in production.
§ 01 — Stack
Production systems that retrieve, reason, and respond. Built around real users, not benchmarks.
Full-stack TypeScript and Python. APIs, dashboards, and the gnarly glue in between.
Cross-platform apps and headless workflows that run themselves at 3am so I don't have to.
Deployed across Azure and a few others. Microsoft AI-102 certified — not just LARPing.
I can read a Figma file, but I'd rather build the prototype in code by lunchtime.
What's on the workbench this quarter — half of these will end up on the next version.
§ 02 — Selected Work
Client work, in production, paying the bills.
Production RAG pipeline indexing millions of unstructured documents for AI-assisted research.
Multi-agent research workflow that autonomously gathers, synthesises, and reports on complex topics.
§ 03 — Writing
A deep dive into building a production RAG system using Databricks, LangChain, and Azure — the architecture, the pitfalls, and what I learned.
I take on 2 retainer clients at a time + occasional fixed-scope builds. Specialty: RAG that ships, agents that don't hallucinate themselves into a corner.