Justin Hong

Education

BS in Artificial Intelligence and Decision Making (Course 6-4) · Expected December 2028

Coursework: Intro to Programming and CS (6.1000), Math for Computer Science (6.1200), Linear Algebra (18.06); Fall 2026: Probability & Inference (6.3800), Fundamentals of Programming II (6.1010)
Activities: MIT Undergraduate Research Opportunities Program (UROP), WMBR 88.1FM Radio, MIT Live

Founding Engineer · March 2026 – Present

First employee (engineer #1 alongside two cofounders) at a YC S26 AI-agent startup whose product has landed deals with a Fortune 10 company
Sole builder of the entire product and web surface: 240+ TypeScript/React components including 170+ interactive agent-workflow demos across 6 industries (finance, healthcare, insurance, ERP, support, accounting)
Own everything from product logic to design system: shipped the full site, interactive demos, and brand used to close enterprise and investor conversations

Squad Leader & Personnel Administration Specialist · July 2024 – Jan 2026

Led a squad in a top-ranked infantry platoon (1st of 48) in CQB room-clearing competition; completed 19 months of national service

Frontend Developer · December 2023 – February 2024

Built a server-side rendered digital exhibition platform (Next.js) with MIT History faculty; archive of activist murals reached 1,000+ users

2026

Built a daemon-based WebKit browser in Rust that makes visual verification nearly free for coding agents: 13 ms median window spawn (p90 35 ms), 216 ms full verify loop; 75+ GitHub stars
Designed the agent-facing control plane: JSON-over-Unix-socket IPC protocol, MCP server over stdio, headless rendering, and human hand-off into a tiling window manager
Sole author across a 3-crate Rust workspace; 350+ commits with published engineering writeups and benchmarks

May 2026

Built an autonomous freight-exception agent in a 2-hour sprint: plan-approve-execute UX on Cloudflare Workers with live SSE trace drafting carrier and customer messages

May 2026

Built an ACL-injury triage dashboard: pose extraction (MediaPipe) + retrieval-grounded rehab guidance turning short video intake into severity bands for 60+ athlete rosters

Systems: Rust, Linux/Wayland, Unix IPC, daemon architecture, WebKitGTK
AI Agent Infrastructure: MCP servers, agent memory systems, LLM tool design, RAG-grounded retrieval
Web: TypeScript, Next.js, React, Python, FastAPI, Cloudflare Workers, AWS, Vercel