rahul@ucla:~$ whoami
cs @ ucla
swe intern @ sagitec
swe @ daily bruindabbling in building ai agents to make my life chill. but the hard (and fun) part is getting them to stop making stuff up and do work efficiently. love seeing my pc handle a lot of mundane things i otherwise would have to do.
- sagitec: building the coding and testing agents in a multi-agent dev pipeline. the catch is they write code in a proprietary low-code language no model has ever seen, so a lot of it is RAG (lucene), MCP tooling, and tuning the agent loop. got token usage per task down 90%+
- opensrcer: open source MCP framework for agent tool-calling. won best use of AI/ML at citrushack (solo) AND presented at GitHub Universe '26. pre-register now at opensrced.vercel.app
- sports analytics forecaster: langgraph agents on top of postgres that look for mispriced player props. the llm runs the queries, it doesn't make up the numbers
- quantitative backtester: my own engine for testing futures / options trading strategies with real transaction costs and walk-forward / out-of-sample testing, custom optionsflow engine
- daily bruin: web stuff for the paper, ~176k readers a month
- won the congressional app challenge with wealthwise and presented it at the capitol
- got a fix merged into worldmonitor (getting into open source)
languages python, typescript, c#, c++, java, sql
ai langgraph, langchain, llamaindex, mcp, rag, pinecone, azure openai
backend fastapi, flask, node, .net, postgres, mongodb, supabase, neo4j
frontend react, next.js
tools git, docker, azure devops
lifting, staring at nasdaq and megacap order flow way too much, watching the niners and kings (unfortunately), playing football
aspiring kitesurfer