Home › Projects
MY PROJECTS
APEXFormula 1 Telemetry OS
Next.js · React Three Fiber · D3A race-engineer's telemetry OS that turns raw Formula 1 lap data into a 3D track ribbon, scrub-synced D3 charts, and exportable generative fingerprints — with ten hand-baked pole laps (2018–2025) and a live OpenF1 next-session countdown.
tap to flipBEAT LABAudio-Reactive Comic Press
Browser Music Player · Canvas VisualizerA fully client-side music player that renders every track as halftone comic art: misregistered ink plates, panel cuts and ink dancers on the beat. It analyses each song once on-device (beats, sections, drops, key) so the stage anticipates the music, hard-cutting to a new “dimension” on the exact beat of a drop, beat-matches tracks like a DJ, and records a 9:16 clip of the drop to share.
Fincil Remastered
AI Finance PlatformA rebuilt Next.js 15 platform featuring an AI Financial Council that debates purchase decisions via live streaming, backed by deterministic financial math, Supabase auth, and pgvector-based retrieval-augmented generation.
tap to flipDinoMem
Distributed AI Memory Layer · Multi-AgentA Postgres-native distributed memory layer for multi-agent systems — hybrid retrieval (semantic + BM25 + entity-graph), Gemini re-ranking, CRDT conflict resolution for concurrent agent writes, bi-temporal windowing for time-travel queries, and receipt-stamped provenance on every operation. TypeScript and Python SDKs with BYOK encryption and HMAC-signed webhooks.
DinoMem
Links will appear on launch
Chatly
Multi-tenant SaaS · RAGAn AI-powered multi-tenant SaaS workspace that ingests private knowledge and queries it through a streaming RAG chat interface, with hard tenant isolation, three-tier RBAC, and Auth0 Organizations for B2B auth.
Chatly
Links will appear on launch
OSS Chatbotwith RAG
Retrieval-Augmented GenerationA web-based chatbot that leverages Retrieval-Augmented Generation to provide accurate, context-aware responses grounded in open-source documents.
