03Playground
Models that run on your device.
No servers, no API keys, no uploads. Each demo downloads a small model once, caches it, and runs inference locally with WebAssembly or WebGPU. Latency numbers are measured live on your machine.
Live demo
What's in this satellite tile?
A ResNet-18 I fine-tuned on Sentinel-2 imagery classifies land cover into 10 classes and shows where it's looking.
- model
- ResNet-18 · int8
- size
- ~11 MB
- runtime
- ONNX Runtime Web
Model in training
This demo goes live once the model finishes training and evaluation. Check back soon.
Live demo
Search my work by meaning
Every project and experience bullet is embedded as a vector. Your query is too, and results are ranked by cosine similarity, so matches don't need shared keywords.
- model
- all-MiniLM-L6-v2 · q8
- size
- ~23 MB
- runtime
- Transformers.js
Loads a small sentence-embedding model (~23 MB, cached after the first visit).
Live demo
Real-time face detection
The detection stage of my attendance system, rebuilt for the browser. It finds faces and landmarks in your webcam feed. It does not identify anyone.
- model
- BlazeFace (short range)
- size
- ~230 KB
- runtime
- MediaPipe · GPU
Uses your camera locally. Frames are processed on your device and never uploaded or stored.
- Faces
- —
- Inference
- —
- FPS
- —