Software Engineer
Ezzy Rappeport
I build AI systems, developer tools, and product software.
Summary
I care about multi-agent systems, humane AI, and tools that compound impact rather than extract attention. My work lives between research and production: turning promising ideas into inspectable, dependable software.
Selected work
MONKEYCLAW
2026Lead Engineer & Contributor
The repo ships a working CLI, seeded demo, live multi-panel dashboard on :8787, eight verifier gates, attack coverage tracking, Telegram alert paths, and a regression model that treats silent controls as incomplete defenses.
18 attack-surface zones · 8 verifier gates · 1,000+ tests
ETCH
2026Founder & Engineer
The saved vertical slice carries a synchronous FIFO through typed specs, three candidates, independent oracle artifacts, simulation and bounded-formal gates, Yosys 0.66 metrics, explicit missing-tool states, and Markdown/JSON proof dossiers. Physical signoff remains pending.
A proven · 50-cycle sim · BMC depth 32 · Yosys 0.66
FLOWE
2026Founder & CEO
The repo has a generated Xcode project, production-shaped config boundaries, local Convex workflows, Canvas sync validation, authentication paths, widgets and Live Activities targets, and simulator build commands for the current iOS app.
SwiftUI · Convex · Canvas · offline retry
VELOX
2026Co-Founder & CTO
The local app ships as a hackathon-ready alpha (v0.3.0) with predictive intent, visible agent tabs, live browsing evidence, cited answer generation, Markdown/JSON export, persisted session snapshots, and a one-command macOS boot path.
Visible agent tabs · no-key demo · Markdown and JSON export
ARGYPH
2026Creator & Engineer
Argyph ships as a read-only MCP server and CLI through npm, crates.io, Homebrew, and release binaries, with incremental indexing and nineteen tools across retrieval, symbols, packing, and local memory.
19 read-only tools · tiered local index · npm, crates.io, and Homebrew
NEXARAD
2026Founder & Engineer
The founder-ready walkthrough can boot the full local demo (`make up`), seed visible research-only studies and draft reports, rehearse uploads, and keep AI providers and PHI disabled by default (`ALLOW_PHI=false`).
Synthetic DICOM demo · PHI disabled · external AI off by default
MATHPILOT
2026Creator & Engineer
The repository includes 613 curated Calc 1/2 problems (production banks around 600-760 per course), a native macOS build, diagnostic and practice loops, mastery and review systems, local homework analysis, ~246 unit tests, and ~26 end-to-end tests across 7 Playwright specs.
613 curated problems · ~246 unit tests · ~26 end-to-end tests
Skills
- Multi-agent systems
- TypeScript / React / Next.js
- Python
- WebGL / GLSL
- Systems design
- Verification & evals
- Product engineering
- Founder / 0→1
Working principles
- Systems over demosI care about the runtime, the failure modes, and the path from a promising prototype to dependable software.
- Evidence over theaterThe strongest product story is a working system with visible constraints, measured outcomes, and honest boundaries.
- Human agency firstAI should make people more capable without making the decisions, provenance, or consequences harder to inspect.