I
Project I · ML infrastructure
Distributed ML Experiment Platform
Runs ML experiments as queued jobs, built to survive repeated submissions and worker crashes without corrupting results.
BUILT WITH
- Python
- FastAPI
- PostgreSQL
- Redis
- Celery
- Docker
View on GitHub ↗
II
Project II · GPU computing
Custom Triton Kernels for GNNs
Fused GPU kernels for the sparse message passing inside graph neural networks, to cut memory traffic and launch overhead.
View on GitHub ↗
III
Project III · Applied AI
Creator Knowledge Base & RAG
Turns YouTube captions and transcripts into searchable knowledge bases, up to 4,000 videos per creator.
View on GitHub ↗
IV
Project IV · Systems
Redis-Style Key-Value Store
An in-memory store built from the socket up, with its own memory pool and a multithreaded TCP path for SET and GET.
BUILT WITH
- C++
- TCP sockets
- Multithreading
View on GitHub ↗