Labs · Eight ready to try

Eight AI systems I built, trained and shipped.

Each one runs for real, in your browser or on a small server I run. Start with an example, see what happens, then open the engineering behind it. No sign-up needed.

Browse the labs

The labs

Pick one and try it.

Each opens on a short overview, then the lab itself. Start with an example; bring your own input when you are ready.

Not sure where to start?

Take the tour, or pick a problem.

Try an application, discover how it works and see the engineering behind it. Examples are practice scenarios, not claims of client results.

New here?

Take the business tour.

Read a receipt. Investigate a pump. Catch a changed answer. Three hands-on examples, with explanations along the way.

  1. Documents
  2. Machine monitoring
  3. Private verification
Start the guided tour

Looking for technical depth?

Open a pipeline, inspect measured results, explore implementation decisions and see the limits.

Explore the engineering ↗
Build status, technology and how to read the numbers
Public
8 of 8 ready to try
Build stages
22 of 31 done
In your browser
4 of 8 need no server at all
  • Lab 01 · Live Vision 3 of 3 stages done

    1. Detection pipeline: Done
    2. Inspector and live eval: Done
    3. Device benchmarks and sign-off: Done
  • Lab 02 · Document Intelligence 2 of 4 stages done

    1. Synthetic invoice data: Done
    2. Fine-tuned model: Done
    3. Extraction pipeline: In progress
    4. Inspector and evaluation: Next
  • Lab 03 · Visual RAG 3 of 4 stages done

    1. Corpus and vector index: Done
    2. Private search: Done
    3. Cited answers in the browser: Done
    4. Vector-space view and evaluation: In progress
  • Lab 04 · Contract Auditor 3 of 4 stages done

    1. Vulnerable samples: Done
    2. Sandboxed audit pipeline: Done
    3. Verified fixes: Done
    4. Editor UI and evaluation: In progress
  • Lab 05 · Zero-Knowledge Inference 3 of 4 stages done

    1. Model and fixed-point export: Done
    2. Circuit and trusted setup: Done
    3. On-chain verifier: Done
    4. Browser prover and evaluation: In progress
  • Lab 06 · Predictive Maintenance 2 of 4 stages done

    1. Pump simulator: Done
    2. Trained anomaly models: Done
    3. Live charts and explanations: In progress
    4. Evaluation: In progress
  • Lab 07 · DeFi Agent 3 of 4 stages done

    1. Market data tools: Done
    2. Mainnet-fork simulation: Done
    3. In-browser planner: Done
    4. Trace UI and evaluation: In progress
  • Lab 08 · Voice Intelligence 3 of 4 stages done

    1. Decoder spike: Done
    2. Streaming transcription: Done
    3. Confidence and word mood: Done
    4. Device benchmarks and sign-off: In progress

How to read the evidence

Live

Measured on this run, on this device, right now.

Live-evaluated

You press Run evaluation and a labelled sample goes through the real pipeline in front of you.

Benchmarked

A full local run, committed to the repository with its provenance.