Jason Liu¶
Founder of 567 Studios. Creator of Instructor (6M+ monthly downloads). Angel investor and a16z scout. Based in New York.
Previously Staff ML Engineer at Stitch Fix (2018-2023). RSI forced me to pivot from coding to teaching. Now I help teams work more scientifically: measuring improvements, iterating quickly, and focusing on what truly matters.
6M+ monthly downloads | 1000+ engineers trained | 11k+ GitHub stars
Subscribe to Newsletter Work with Me
What I Believe¶
- Your AI problems aren't AI problems. They're usually data, process, or people problems.
- Sell outcomes, not tasks. Measure what matters: does this solve a real problem someone will pay for?
- Simplicity beats complexity. If you can't explain it simply, you don't understand it.
- Experimentation speed is your only moat. Confidence is the memory of success; practice more than you play.
Consulting¶
I help companies turn prototypes into production systems. No maintenance contracts. No dependencies. Just results.
| Company | Contact |
|---|---|
| Zapier | VP of Product |
| HubSpot | GM |
| Limitless AI | CTO |
| Weights & Biases | CTO |
| Modal Labs | CEO |
| Pydantic | Creator |
| Timescale | VP Product |
| Raycast | Team |
Also: Enterpret, Tensorlake, Trunk Tools, Naro, New Computer, Sandbar, Dunbar, Bytebot, Kay.ai, Galileo, Distributional
Training¶
Systematically Improve Your RAG¶
A 6-week certification for engineers who want to build production RAG systems. 4.8 rating from 1000+ engineers.
Where students come from: OpenAI, Anthropic, Google, Microsoft, Amazon, Airbnb, Databricks, TikTok, Shopify, Nvidia, Meta, GitLab, Intercom, PostHog, Miro, Workday, Mozilla, Red Hat, Gumroad, Visa, PwC, KPMG, Accenture, Bain & Company, Booz Allen Hamilton, Northrop Grumman, DataStax, Decagon
Free 6-Day Crash Course Learn More
AI Coding Accelerator¶
A 3-day intensive with Vignesh Mohankumar on building with AI tools.
What People Say¶
"As an Applied AI Engineer at Anthropic, I was familiar with all of the standard retrieval methods and RAG papers, but Jason's frameworks helped me operationalize what I'd learned and it's had an incredibly positive impact in my work with customers."
"Jason gave our AI product the horsepower it needed. He's a pro if you need real results fast."
"Full-stack AI-business guidance in a box. The value went beyond just product acceleration."
"Brought a wealth of experience and understanding that helped us develop better ideas and move faster. 100% worth it."
Open Source¶
I created Instructor, a Python library for structured outputs from LLMs. OpenAI cited it as inspiration for their structured output feature.
11,000+ stars on GitHub | 6M+ monthly downloads
Investing¶
Angel investor and a16z scout. Working relationships with Weekend Fund and Script Capital.
- Pydantic Logfire - $12.5M Series A, Sequoia
- Haize Labs - $100M valuation
- Browserbase - $44.4M
- Exa - $22M, YC 2021
- Lovable.dev - $20M+
- Extend - $17M Series A
- Modal - $16M Series A
- Julius AI - $11M
- Shaped AI - $8M Series A, YC 2022
- Rork - $2.8M seed, a16z Speedrun
- Godel Terminal - $2M pre-seed
- Kino.ai - YC S23
- Raindrop.ai - YC
- Sandbar - Seed, True Ventures
- Augmental - $1M+
- Daytona
- Vantager
- Argo Labs
- Smithery.ai
- Empower Sleep
Background¶
| Stitch Fix | Staff ML Engineer (2018-2023) |
| Meta | Data Scientist (2017) |
| South Park Commons | Sabbatical |
| University of Waterloo | Computational Mathematics & Statistics |
Connect¶
Consulting: [email protected]