·
Builders & Technical Leaders
Jason Liu
Creator & Founder |Instructor / 567 Studios
Created Instructor (structured LLM outputs via Pydantic), cited by OpenAI as inspiration for their structured output feature. AI consultant and educator.
Biography
Deep Research
Jason Liu is a staff-level machine learning engineer, angel investor, a16z scout, and the creator of Instructor, a Python library for structured outputs from LLMs with 12,500+ GitHub stars and 6M+ monthly downloads. The OpenAI team cited Instructor as direct inspiration for their native structured output feature. Liu studied Computational Mathematics & Statistics at the University of Waterloo (2012-2017). He worked as a Data Scientist at Meta (Facebook) in 2017 on content detection algorithms at 2B+ user scale, then spent five years as a Staff ML Engineer at Stitch Fix (2018-2023), where he led a team of 6-7 engineers building multimodal embedding systems and created the Flight framework processing 350M+ daily requests. In 2023 he founded 567 Studios, a solo AI consulting practice advising seed-to-Series B startups including Zapier, HubSpot, Limitless AI, Weights & Biases, Modal Labs, Timescale, and Pydantic. He ran cohort-based training programs on Maven with students from OpenAI, Anthropic, Google, Microsoft, Amazon, and McKinsey. In February 2026 he sunset 567 Labs and open-sourced all course content. He is based in New York, where he works as a freelance ML consultant and angel investor.
Structured LLM OutputsPydantic IntegrationRAG SystemsContext EngineeringAI ConsultingMultimodal EmbeddingsEvaluation FrameworksFunction CallingOpen SourceAI Engineering Education
Timeline
12 Research12 total
2026
2025
2024
2023
2018
2017
2012
Key Contributions
Instructor
Python library for structured outputs from LLMs, patching provider SDKs to return Pydantic models. 12,500+ GitHub stars, 6M+ monthly downloads, cited by OpenAI as inspiration for their structured output feature.
Instructor JS/TS
TypeScript port of Instructor for structured extraction from LLMs in the JavaScript ecosystem.
Flight Framework (Stitch Fix)
Internal ML pipeline framework at Stitch Fix processing 350M+ daily requests with 80% internal adoption, serving as a semantic bridge integrating multiple systems.
Systematically Improving RAG (Maven Course)
6-week hands-on course covering synthetic evaluation, embedding fine-tuning for 20-40% gains, query segmentation, and multimodal indices. Students from OpenAI, Anthropic, Google, Microsoft, Amazon.
567 Labs Open-Source Course Archive
Open-sourced written content from RAG Playbook and Consulting courses as freely available ebooks after sunsetting 567 Labs.
Context Engineering for RAG
Prolific writing on building better agentic RAG systems, evaluation frameworks, and practical AI engineering patterns.
Notable Quotes
“
It's just Python, right? Like, if you're going to use the LLM SDKs, you're obviously going to install instructor.
“
Really, I care a lot more than just the fact that it is JSON. Function calling really shines when you can specify the schema.
“
I just don't think it's a billion dollar company. I think if I make a million dollars as a consultant, I'm super happy.
“
Agency is just a matter of having courage and doing the thing that's scary. The higher agency person just falls down and tries.
“
Most of the time you're not going to need someone writing PyTorch code. You're gonna need someone to write instructor code.
12 sources(click to expand)
Jason Liu - personal website & introductionsJason Liu - Resumejxnl (Jason Liu) - GitHub profileInstructor - structured outputs for LLMs (GitHub)Instructor - Python documentation siteHigh Agency Pydantic > VC Backed Frameworks - Latent Space podcastJason Liu - Instructor, Shipping LLMs to Production - devtools.fmInstructor with Jason Liu - Weaviate Podcast #88Systematically Improving RAG Applications - Maven courseSunsetting 567 Labs and Open Sourcing the Course ContentBridging Language Models with Python - Jason Liu on MediumInstructor on PyPI
Research generated March 19, 2026
Builders & Technical Leaders/Jason Liu
All Profiles