Principal Engineer / Architect — GenAI, LLM agents, RAG. Hyderabad, India.
I build AI systems that run in production: multi-step agents with tool calling, retrieval grounded in a customer's own content, and the evaluation and observability that make non-deterministic behaviour trustworthy. 18+ years in software — .NET/Azure and TypeScript, the last four in applied GenAI — and I have also built and led the teams that ship it.
prashant-anand.github.io · LinkedIn · prashantanand@outlook.com
Most of my production work lives in the @samisen-io organisation (my consultancy). These are the five repositories I would want a reviewer to read first — each has a full README stating what works, what the tests cover, and what is deliberately missing.
| Repository | What it is |
|---|---|
| samisen-io/venue-assistant | MCP server in C#/.NET 10 exposing venue and event operations as agent-callable tools — vendor matching, budgets, timelines, risk — with the decision logic in unit-tested calculators · 10 tools · xUnit |
| samisen-io/eventonaut-api | The RAG engine behind two conference apps: LangChain over Pinecone with one namespace per conference, Redis/Upstash conversational memory, streamed answers, per-call token accounting · Python, FastAPI |
| samisen-io/venue-assistant-saas | Multi-tenant venue-management SaaS — tenant isolation enforced in the database with RLS, 83 API routes, an AI layer behind one Claude interface, 31 Playwright specs · Next.js, TypeScript, Supabase |
| samisen-io/mcp-server-web | ASP.NET Core MCP server giving agents a typed file API over Azure Blob Storage — list, upload, download, delete, exists · .NET 8 |
| samisen-io/aio_ready | Audits a page for how AI answer engines read it, built on Claude · Node.js, Express |
On this account:
| Repository | What it is |
|---|---|
| docagent | Agent that turns raw documents (PDF/DOCX/TXT/Excel) into structured data · Python (phase 1a — the README is honest about what does not work yet) |
| unicoach | Profile-building platform for university applications — activity tracking, evidence, AI-matched opportunities · Next.js, Supabase, Claude |
| customer-analytics | Customer-analytics probability models (NBD) shipped as an Excel add-in · C#, Excel-DNA (academic project at WashU) |
| prashant-anand.github.io | This site — experience, selected work, and both resume variants for download |
- Deterministic core, model at the edge. Arithmetic a user might dispute belongs in tested code, not in a prompt.
- State the limitation. Every README here carries a "limitations and next steps" section — no eval harness yet, no browser tests, single-region. A portfolio without one is a brochure.
- Evaluation before scale. Agents ship with tracing, cost and latency tracking, and regression evals against a curated set — not vibes.
LangChain/LangGraph · LangSmith · Pinecone · MCP · Azure AI Studio · C#/.NET · TypeScript, Next.js, React Native · Python · AWS EKS, Azure · Supabase/Postgres · PCI DSS, SOC 2