a few course, research, hackathon, and independent projects. reports, demos, presentations, and writeups are linked where they are available.
python · polars · pandas · networkx · statsmodels · scikit-learn · anilist graphql
Built a 2000-2024 anime panel from AniList and MyAnimeList/Kaggle data to test whether title success clusters by studio identity or by repeated creative staff. The pipeline compares studio graphs, staff-projection graphs, permutation tests, and fixed-effect regressions.
typescript · next.js · react · three.js · spark · world labs api · gemini · express · zustand
HackPrinceton Spring 2026: Best AI-Powered App Sponsor Track Award
Built an AI-powered 3D history learning app that turns natural language prompts into explorable scenes, with a React interface, Gaussian-splat viewer, conversational narrator, text-to-speech, and World Labs world generation.
typescript · next.js · react · solidity · foundry · rust · axum · postgres · viem · wagmi
MIT Bitcoin Hackathon 2026: People's Choice Award
Built a full-stack crisis relief platform with donor and NGO flows, Solidity donation vaults, a Rust event indexer, Postgres-backed API services, and a globe UI for crisis regions on the Humanity testnet.
python · pandas · scikit-learn · tensorflow · keras · pyarrow · statsmodels
Modeled October 2025 NYC yellow and green taxi trips to predict ride duration and whether a passenger would tip before the trip ended. Compared random forests, logistic regression, and feed-forward neural networks after cleaning parquet trip records and handling skewed durations.
python · streamlit · csv logging · pomdp design
Designed and implemented a Streamlit experiment for a structured text-based conversation task with a simulated partner, studying disclosure as a sequential decision under uncertainty. The app randomized timing and reciprocity conditions and logged turn-level responses for analysis.
python · tensorflow · scikit-learn · pandas · praw · tweepy · ccxt · textblob · stable-baselines3
Built a sentiment-driven crypto allocation pipeline using Reddit, Twitter, and market data, LSTM-based SHIB return forecasts, volatility models, and portfolio-allocation logic for SHIB-USDC trading under different risk preferences.
more coming as i finish things.