Projects

Finance Research Agent

An AI agent that autonomously researches stocks by combining real-time market data, semantic search over SEC 10-K filings (RAG), and web research via a specialized subagent.

Claude Agent SDK RAG MCP Pinecone Python
Finance Research Agent Architecture
  • 10 custom MCP tools — 6 finance data + 3 search modes (naive, hybrid, agentic) + SEC filing search
  • RAG pipeline — 4,575 SEC filing chunks in Pinecone serverless, section-aware chunking, local embeddings, metadata-filtered retrieval
  • 3 search modes — naive vector (simple), hybrid (vector + keyword), agentic (multi-step decomposition). Agent chooses the right mode autonomously.
  • Multi-agent — research subagent delegates web search to a separate context
  • ~700 lines of Python, 3 dependencies
Patterns: tool use, RAG, multi-agent, MCP, async streaming View on GitHub →

Automated Trading Platform

Systematic trading system with multi-factor stock scoring, regime detection, AI-powered thesis generation, and automated order deployment with human-in-the-loop approval.

Python Claude API SQLite
Automated Trading Platform Architecture
  • 5-pillar factor model — valuation, quality, momentum, risk, forward-looking with walk-forward validation
  • Regime scoring — adjusts positioning based on macro conditions
  • Narrative risk overlay — maps portfolio exposure to macro themes
  • SEC EDGAR pipeline — pulls 10-K filings, extracts key sections, feeds AI analysis
  • Automated deployment — AI generates pitches → quality gate → order sheet → human approval → execution
Private repo — available for discussion in interviews

Health Analytics Platform

Local-first health system integrating Apple Health data with AI-powered journaling, habit tracking, and weekly reviews. All data stays on-device.

Python SQLite Claude Code
  • Apple Health ingestion — workouts, heart rate, sleep, activity metrics
  • AI-powered briefings — morning/evening/weekly reviews synthesizing health data with journaling
  • Habit tracking — 7 categories with streak monitoring and trend analysis
  • Privacy-first — all data local, no cloud services, no syncing
Private repo — architecture available for discussion

Interactive Learning Knowledge Base

49-guide knowledge base covering system design, data structures, AI engineering, security, nutrition, and more. Built as self-contained HTML with dark theme and interactive features.

Python HTML/CSS/JS
  • 49 guides — from options pricing math to Claude Code anti-patterns
  • Interactive features — skill trackers, cost calculators, streak counters
  • Automated sidebar — Python script manages navigation across all pages
  • Topics span — tech fundamentals, life optimization, tools & workflows
Private — selected guides available for sharing