Architected a privacy-first personal AI assistant and custom language framework designed to optimize local-first AI code synthesis and low-latency inference.
Engineered for Intelligence,
Built for Scale.
"Focused on privacy-first local AI systems, mechanistic interpretability from first principles, and robust full-stack software."
Mechanistic interpretability framework utilizing PyTorch and TransformerLens to extract internal model activations and train layer-wise probes for safety guardrails.
Gamified task planner and Pomodoro application featuring customizable task durations, countdown timers, and automated DNS routing via Cloudflare.