I've previosuly founded two companies: a blockchain signals intelligence startup, and a CRM for teams selling to Web3 companies. Not very successful; learned a lot.

Since, I've helped a handful of startups with their fundraising, including Diode and Sea12.

- Conflicting Expectations: Convergence and Chaos in Simple High-Dimensional Games

- Cyclical Electromechanical Error Denial System Using Matrix Profile

- Solving IVPs with Physics-Informed-Neural-Networks

- LLM Numerical Bias Investigation

- VectorBridge: A Robust Technique for Cross-Model Embeddings Translation

In varying degrees of completion:

- GameNgen Open Source Repro.

- LLM Numeric Bias Investigation.

- CommaVQ Open Source Repro.

- Physics-Informed-Neural-Network Demos.

- PreImage - CLI tool for Python dependency optimization.

- EdgeEmbed: Embeddings models run at the network edge for lowest-latency vector search.

- JSGrad - Micrograd rewritten into Javascript for edge-ML.

- This website.

- World models, neural simulation, and model-based RL.

- National security, defense technology, and OSINT.

- Lightweight ML: edge inference, model quantization, digital signals processing.

- Controls, planning, and mechatronics engineering.

- Product, interface, and interior design (Bauhaus, Ceramics, Dieter Rams).