Michael Yang
CS at Columbia, class of 2030.
AI evaluation and robotics.
Projects click to expand
Coding agent memoryMemory benchmarks and a RAG alternative
At Columbia DAPLab, I’m working on benchmarks for memory in coding agents and an alternative to retrieval-augmented generation.
In progress. Planned NAACL 2027 submission.
Mortgage benchmarkEvaluating agents on mortgage workflows
Built a mortgage origination benchmark at a stealth AI startup with a custom simulation environment, 50+ agent tools, and a versioned HTTP API. Combines sourced data, synthetic loan files, and controlled mutations to test document review, borrower follow-ups, and product eligibility. Automated graders evaluate complete action traces for workflow correctness and evidence-supported financial claims.
XAIGID-RewardBenchTesting how models judge image explanations
FTC roboticsRobot hardware, controls, and path planning
2022–2026. Led software, hardware, and manufacturing for two seasons. Built custom path planning, computer vision, and countersprung lift mechanisms; also worked on funding and sponsorships.
Three World Championship qualifications and a World Championship Inspire Award with my team. Official Team USA alumnus.
Water polo scoutingPlayer tracking and game observation
Led an AI Club project using machine learning and computer vision for player tracking, game observation, and strategy support. Founded and led the 30+ member club from 2024 to 2026.
BackgroundEducation, experience, skills
Education
Columbia University SEAS · B.S. in Computer Science, expected 2030. C.P. Davis Scholar.
St. Mark’s School of Texas · 2022–2026.
Experience
- Columbia DAPLab
- Undergraduate Student Researcher · Sep 2026–present
- Stealth AI Startup
- Researcher · Sep 2026–present
- Stealth Startup
- Founder · Jun 2026–present. Selected for Wix Ventures’ Foundry workshop.
- UT Dallas
- Researcher · Sep 2024–Aug 2026
- HStar Space
- Summer Intern · May–Aug 2025
Leadership
Robotics founder and technical lead, 2022–2026. AI Club founder, 2024–2026.
Skills
Python, Java, JavaScript/TypeScript, C++, SQL. PyTorch, Hugging Face, computer vision, Onshape, controls, and manufacturing.