projects/

Projects

What I've built. The two big ones first.

PitWall

F1 Strategy Intelligence Platform

A full-stack F1 strategy platform built on FastF1 and OpenF1 data. The Strategy Replayer maps stints, lap times, tyre degradation and undercut windows; Driver DNA fingerprints each driver; the Championship Engine runs Elo updates through 50,000-run Monte Carlo simulations; and a Race Predictor and Live Tracker cover race weekends. Team Radio Sentiment uses the Claude API to read the mood on the radio.

Next.js
TypeScript
D3.js
Tailwind
Python
FastAPI
FastF1
OpenF1
Claude API
Supabase

Live Code

SpaceMetrics

Soccer CV Analytics

in progress

An end-to-end computer vision pipeline for soccer match footage. A YOLOv8 detector fine-tuned on SoccerNet tracking data reaches 0.955 mAP50 on players and referees, a tuned ByteTrack setup cut identity fragmentation from 133 to 80 track IDs for about 25 players, and jersey-colour k-means separates the teams. Homography pitch calibration, accurate to under 1m, turns pixels into real positions for zone-density stats.

Python
PyTorch
YOLOv8
ByteTrack
OpenCV
scikit-learn
pandas

Code

More projects

CryptoSentinel
A real-time crypto monitor that pairs live prices with VADER sentiment scores on news headlines, and fires alerts when the two line up, like a price drop plus bearish news. Alerts go out through Discord, a CLI and an auto-refreshing React dashboard, which also tracks portfolio P&L.
Python
Flask
NLP (VADER)
React
TypeScript
Code for CryptoSentinel on GitHub (opens in a new tab)
From Raw to Ready
A production-style ML data pipeline that turns messy raw CSVs into validated, versioned, model-ready features. Raw rows are stored as immutable JSON events in PostgreSQL, validation fails fast with a machine-readable report, and SQL builds rolling features with time-aware train/val/test splits and a feature registry for lineage.
Python
pandas
SQL
PostgreSQL
Docker
Code for From Raw to Ready on GitHub (opens in a new tab)
Transfer Equity Predictor
Predicts which student groups are at risk of below-average UC/CSU transfer rates across 15 California community colleges, using official CCCCO outcome data from 2015–2022. A Random Forest trained on 847 records reaches 83% accuracy, with SHAP explanations and a Streamlit dashboard for exploring the gaps. This one's personal: I'm a De Anza transfer myself.
Python
pandas
scikit-learn
Streamlit
Plotly
SHAP
Code for Transfer Equity Predictor on GitHub (opens in a new tab)
Zero-Waste ML
Flags U.S. industrial facilities at risk of missing zero-waste targets (a diversion rate under 90%) using EPA Toxics Release Inventory data. A Random Forest trained on 88,000+ facility records from 2020–2023, with SHAP to surface the top predictors and a Streamlit dashboard with a Plotly map.
Python
pandas
scikit-learn
Streamlit
Plotly
SHAP
Code for Zero-Waste ML on GitHub (opens in a new tab)
TEDx De Anza website
The event website for the TEDx club at De Anza College.
React
TypeScript
Vite
Tailwind
Live site for TEDx De Anza website (opens in a new tab)Code for TEDx De Anza website on GitHub (opens in a new tab)

Projects

PitWall

SpaceMetrics

More projects

CryptoSentinel

From Raw to Ready

Transfer Equity Predictor

Zero-Waste ML

TEDx De Anza website