Data Scientist at Zurich North America with 5 years of experience shipping production ML — and, increasingly, production LLM/RAG and AI-agent systems. I just finished my Master of Computer Science (MCS) at UIUC. Based in Chicago; interested in data science, ML/AI engineering, and distributed-systems roles.
- 🔭 At work: building a retrieval-augmented (RAG) document-understanding agent for underwriting; previously rebuilt production pricing models (Tweedie GLMs, PySpark/Databricks) used by underwriters.
- 🧠 Focus: LLM applications, RAG, AI agents, distributed systems, and turning models into things people actually use.
| Project | What it is | Stack |
|---|---|---|
| distributed-stream-processor | Distributed stream-processing system from scratch — SWIM failure detection, consistent-hashing replicated file system, exactly-once streaming; 10-node Docker cluster | Go, Docker |
| ai-agent | Minimal coding-agent harness — tool calling, memory, SWE-bench eval — across multiple LLM providers | Python, LLMs |
| devils-advocate · live demo | Real-time AI debate app that stress-tests startup ideas — voice + text debaters with RAG memory | React, FastAPI, Socket.IO, Gemini |
| kaggle-bench | Benchmark for LLM-agent repair of tabular preprocessing plans, grounded in 20 Kaggle competitions | Python |
| code-llm-prompting-study | Research on LLM code generation & understanding (DeepSeek Coder, HumanEval, multilingual) | Python, PyTorch |
| srsly-fit · live demo | Full-stack workout planner — exercise search, templates, set tracking | Next.js, NextAuth, MySQL |
Languages: Python · Go · SQL · Java · R ML/AI: LLMs · RAG · vector databases · PySpark · pandas · scikit-learn · statsmodels Systems/Cloud: Docker · Azure (Functions, Databricks) · distributed systems · Streamlit · GitHub Actions