Most portfolios tell you what someone can do. For an ML/AI engineer, that's a missed opportunity — you can show it, live, in the browser.
So I rebuilt mine around a simple idea: the site itself should be the demo.
The portfolio agent
The homepage is a chat agent you can ask anything about my work. It runs on a small intent engine over my own data, and you can flip on a free in-browser LLM (WebGPU) for free-form answers — no server, no API key, no cost. The medium is the message: if I claim to build agents, you should be able to talk to one.
Live ML you can touch
Two demos run entirely on your device:
- AI personal trainer — your webcam, pose estimation, and a rep counter that checks squat depth and coaches you out loud. It's my RepRight + VoiceCoach projects, merged.
- Draw-a-digit — a neural net I trained on MNIST (~97% accuracy) running as plain JavaScript. No TensorFlow, no model server — just the weights and a forward pass.
Why it matters
A recruiter who experiences a model believes it instantly — far more than a bullet point. And it's shareable, which is the only real way a portfolio earns repeat visits.
If you want the details on any of these, just ask the agent on the homepage.