Parc Fermé
ML / race predictions
XGBoost model that sets each driver's win, podium, points and DNF odds once qualifying is done, published every race weekend. Since going live it has called 78% of podium finishers, against 56% from the starting grid.
São Paulo GP: 2 of 3 podium spots called, NOR won.
Python, XGBoost, scikit-learn, pandas, FastF1, Jinja2, GitHub Actions
Case studyLive siteSource
VibeCheck
Security tooling
Hybrid SAST + DAST scanner for AI-generated web apps. Flags hardcoded secrets, unauthenticated routes and weak JWT config, then fuzzes the running app for IDOR, missing rate limits and leaky errors.
Python, requests, PyJWT, Rich
Source
ReSole
Mobile / marketplace
Marketplace for buying, selling, swapping and donating used sneakers in the Philippines. Gemini checks each listing's condition photos and matches it to known models by embedding search before it goes live.
One of three developers: I built the Gemini photo pipeline, the six-angle upload, and the schema and migrations.
React Native (Expo), Next.js, PostgreSQL, Supabase, Drizzle, Gemini
Source
wordwork
Offline-first PWA
Daily language study app for German, Italian and Japanese: spaced-repetition flashcards, a der/die/das drill, a study log with streaks and a one-sentence journal. Works offline from IndexedDB and syncs through Supabase.
React, TypeScript, Vite, Dexie, Supabase, PWA
Live siteSource
AnimoNotes
Student forum
Note-sharing forum for DLSU students: rich-text posts with votes and bookmarks, threaded comments, polls, messages and reporting tools.
Led the team and built most of the app.
Next.js, TypeScript, Prisma, PostgreSQL, Better Auth, Tiptap
Source