Projects
AI Stock Analysis Agent
A price prediction without market mood is only half the picture, and a sentiment score without a number is just a guess. So this agent does both: a PyTorch LSTM I trained predicts the next-day price move, while VADER scores the news sentiment in real time. Then Claude merges both signals into one BUY / HOLD / SELL call, so it reflects where the stock might go and how the market currently feels about it.
Financial Intelligence Assistant
For a class project, we wrapped our Financial Intelligence AI system into a Streamlit app. You can chat directly with our agents to get financial insights, and switch between a Baseline and a Multi-Agent architecture to compare them.
| Metric | Q1'26 | Q4'25 | YoY |
|---|---|---|---|
| Current Ratio | 1.5x | 1.4x | +7.1% |
| Debt/Equity | 0.8x | 0.85x | -5.8% |
| ROE | 18.2% | 17.5% | +0.7% |
UW Foster MBA — AI Course Pilot
A RAG-based assistant I built for UW Foster MBA students to help them plan their electives. It searches 49 real course syllabi and recommends classes based on career goals. Because of IP restrictions from Foster, full access only works for enrolled MBA students.
Multi-Agent Chatbot System
A frontend I built on top of a multi-agent architecture from a class project. Users can chat securely with different specialized AI agents, and the system keeps the chat history and moderates the community.