Ezzy Rappeport

Software Engineer / AI Systems / Founder

I build AI systems, developer tools, and product software.

Explore work

Based in California

01 / SELECTED WORK

Selected projects.

Native software and agent systems.
The decisions behind the work.Drag the artwork. Explore the case study.

macOS · Available now

Downright

Markdown, treated like a Mac document.

I built the native document experience, exact-source editing, and Finder integration.

Creator & EngineerInside the editor

Recorded demo fixture

MC-2026-0001

A backup request crossed the sandbox boundary.

Finding confirmedReproduction saved

Trigger: filesystem_breach
Replay: 1 request
Next boundary: No patch recorded

Agent security

MonkeyClaw

Tests agent sandboxes, reproduces failures, and checks that defenses leave evidence.

My contributions: native telemetry ingestion, transcript replay, and the evidence dashboard. Follow one recorded demo finding.

Engineering ContributorFollow the finding

iOS · Planning & focus

FlowE

Coursework, calendars, and focus in one iOS app. Start with a brain dump; leave with a plan for today.

A native SwiftUI build with Canvas sync, real-time tasks, and offline retry.

Founder & CEOView project

Hardware design

Etch

AI-assisted hardware design where generated RTL earns its way through verification.

A saved FIFO design run with simulation, bounded-formal checks, and a proof dossier. Physical signoff pending.

Founder & EngineerView project

Developer tools

Argyph

Useful codebase context for coding agents, through one local MCP server. No cloud account required.

A released Rust CLI with 19 tools, distributed through npm, crates.io, and Homebrew.

Creator & EngineerView project

More things I’ve built

Explore GitHub

Velox

A Chromium browser where research agents work in visible tabs.

NexaRad

Evidence-linked imaging workflows with strict non-clinical demo boundaries.

MathPilot

A calculus study app that connects diagnosis, practice, and review.

TerminusIn development

A coding-agent workspace built around a Rust execution kernel, scoped tools, and verification. Still in development.

CoOpsIn development

Department agents hand work across permission boundaries, with human approvals and traceable receipts.

PlateFowardPrototype

From donation photos to food-rescue offers and real-time recipient matching.

FollowThroughDemo

Meeting notes become clear recaps, assigned actions, and follow-through.

ScoutSSDPrototype

Scholarship discovery, eligibility filtering, and application preparation with human review.

02 / A LITTLE ABOUT ME

Curiosity.
Then code.

I build product interfaces and the systems that make their behavior understandable.

In Downright, that means preserving a person’s source text through reading, editing, and external changes. On MonkeyClaw, my work connected native telemetry, transcript replay, and a dashboard that keeps evidence in view. Those are different products, but the details that preserve trust matter in both.

I’m based between Los Angeles and Santa Cruz, and open to software and AI opportunities.

View résumé