Ezzy Rappeport

Building AI-native products at the intersection of multi-agent systems and real-world deployment. Currently leading Monkeyclaw and advising two seed-stage AI companies on agent architecture and evals.

• Designed and shipped a multi-agent security platform from zero to closed beta in 9 months.
• Advised portfolio companies on agent eval methodology, reducing hallucination-related incidents by 60%.
• Published an open-source agent capability framework adopted by 800+ developers in the first quarter.

TypeScriptPythonLangGraphRustKubernetes

Joined as the second engineer and built the core inference pipeline for a regulatory-cleared medical imaging triage system. Took the model from research prototype to FDA submission.

• Architected the inference pipeline serving 90-second median time-to-triage at 99.95% uptime.
• Led the data engineering effort to ingest and de-identify 2.4M studies across 38 institutions.
• Owned the regulatory ML documentation package that secured FDA 510(k) clearance in 11 months.

PyTorchCUDAC++DICOMKubernetes

Founded and led an adaptive learning company serving community college students. Built the mastery model, raised a pre-seed, and ran a 1,200-student pilot that outperformed control by 0.41σ.

• Raised a $1.8M pre-seed from education-focused investors.
• Designed and ran a 1,200-student randomized pilot with a state community college system.
• Open-sourced the Bayesian mastery model, which has since been forked by 6 university research labs.

TypeScriptReact NativeFastAPIPyTorch