DEDSEC//BROADCAST · CH.07 · LIVE FROM BENGALURU
AI Software Engineer
Python · LLMs · RAG · MCP · React
Adhrit
Verma
Builds multi-agent systems that do real work — contract review, accessibility remediation, schema retrieval for agents — and ran a regulated airline platform solo for two years.
I buildAI systemsthat doreal work.
Multi-agent pipelines. Retrieval that has to actually retrieve. MCP servers that let agents touch a real database safely — each one tested, measured, and honest about its limits.
On the record
3
Years in production
Since Sep 2023
3
Production AI systems
Contrast · ClauseGuard · TableFox
162
Tests across those pipelines
72 + 23 + 67, all passing
−29%
Agent token usage cut
TableFox vs. full-schema baseline
Featured files
// Recovered — research and reference builds
Recovered files
Loadout
Six modules, 35 tools — chosen per problem, never per trend.
- AI / ML / GenAI
- Large Language Models, LangChain, LangGraph, RAG, Hybrid retrieval, MCP, AI agents, Multi-agent orchestration, Prompt engineering
- Languages
- Python, JavaScript, TypeScript, SQL
- Backend
- FastAPI, Node.js, Express.js, REST APIs, RBAC
- Data & Vector
- PostgreSQL, MongoDB, Redis, SQLite, Local embeddings (Ollama), Schema design, Query optimization
- Frontend
- React.js, Next.js, HTML5, CSS
- Cloud & Delivery
- AWS Lambda, S3, DynamoDB, Docker, GitLab CI/CD, Git
Open a channel
Agent pipelines, retrieval that has to actually retrieve, or a platform that has quietly become three platforms — send it over. Response time beats most CI pipelines.
Establish connectionBengaluru, India · Open to remote