I'm an AI engineer who got tired of agent demos that only work in a Jupyter notebook, so now I build the ones that survive real production traffic. MS in Computer Science from UMBC, based near DC, currently on the market and open to my next role.
Most of what I ship lives at the intersection of multi-agent orchestration, RAG, and "how do we know this thing isn't lying to us," which is a nicer way of saying I spend a lot of time red-teaming my own agents just to see how badly I can break them.
const mohan = {
role: "AI Engineer",
based_in: "Vienna, VA",
currently: ["building agents", "job hunting", "breaking my own prompts on purpose"],
favorite_stack: ["LangGraph", "FastAPI", "Qdrant", "Langfuse"],
fun_fact: "shipped an autonomous GitHub bot that opens its own PRs",
};- 🤖 Multi-agent orchestration — LangGraph, hierarchical supervisor patterns, human-in-the-loop approval gates
- 🔍 RAG done properly — hybrid retrieval (BM25 + vectors), reranking, GraphRAG on Neo4j, not just "stuff it in a prompt"
- 📊 Evals that actually mean something — Ragas CI gates, Langfuse tracing, drift detection, because "vibes-based" LLM shipping doesn't scale
- 🛡️ Breaking things on purpose — OWASP LLM Top 10 red-teaming, prompt injection defense, guardrails at agent boundaries
- ⚡ AI-assisted everything — Claude Code and Cursor run alongside me daily, because writing every line by hand in 2026 is a choice, and not the fast one
|
🛡️ LLM Security Sentinel Red-team platform running 6 OWASP attack agents (prompt injection, model DoS, excessive agency) against any LLM endpoint, with a human approval gate and a live Next.js dashboard streaming results over SSE.
|
🧠 ARIA — Autonomous Research Agents 5-agent research platform with CI-gated evals (faithfulness > 0.91 on golden questions), hybrid BM25 + vector retrieval, cross-encoder reranking, and a sub-10ms semantic cache so it's actually fast.
|
|
🕸️ GraphRAG Knowledge Platform Ingests documents, images, and recursive links up to depth 5, builds a Neo4j knowledge graph, and answers questions with full citation tracing instead of confidently making things up.
|
🔁 OSS Automation Agents Two autonomous workers living entirely in GitHub Actions cron jobs: one digests PR/CI activity into email reports, the other writes fixes, runs tests, and opens its own draft PRs. Zero local infra.
|
📌 Repo links go here once these are pushed public, swap the project titles above into actual
github-readme-statspin cards for a live-updating version.
Languages
Agentic / LLM
Backend / Frontend
Data / Infra
Observability / Evals
- M.S. Computer Science — University of Maryland, Baltimore County
- B.S. Computer Science — Jawaharlal Nehru Technological University, Hyderabad
- Certified: Google AI Professional Certificate · Azure AI Engineer · AWS Solutions Architect · Agentforce Innovator · Claude Code in Action
