Jason Liu¶
Staff ML Engineer | AI Consultant | Open Source Creator
[email protected] | jxnl | jxnlco | jxnlco
Former Staff MLE at Stitch Fix. Creator of Instructor (6M+ monthly downloads, 11k+ stars, cited by OpenAI). 10+ years building AI systems at scale. I help companies establish thought leadership in AI, build evaluation frameworks, and drive developer adoption.
Looking for my next 4-5 year project.
Experience¶
Founder, 567 Studios¶
2023 - Present
- Work with Seed to Series B companies on AI best practices: RAG, Context Engineering, Evals
- Consulting clients: Zapier, HubSpot, Limitless, Weights & Biases, Modal Labs, Timescale, Pydantic
- Built training programs on Maven for engineers from OpenAI, Anthropic, Google, Microsoft, Amazon, Netflix
- Designed workflow automations and fine-tuned embedding models across industries
Staff Machine Learning Engineer, Stitch Fix¶
2018 - 2023
- Built multimodal embedding system using ResNet-50 and CLIP+GPT-3 integration
- Designed Flight framework handling 350M+ daily requests with 80% internal adoption
- Built styling chatbots with DaVinci, improving recommendation quality by 40%
- Technical lead for team of 6-7 engineers and data scientists
Data Scientist, Meta (Facebook)¶
2017
- Developed ML algorithms detecting harmful content with 95%+ accuracy at 2B+ user scale
- Built real-time monitoring dashboards, reducing escalation time by 50%
Open Source¶
Instructor - Python library for structured outputs from LLMs
- 6M+ monthly downloads, 11,000+ GitHub stars
- Cited by OpenAI as inspiration for their structured output feature
Education¶
University of Waterloo - Bachelor of Mathematics
Computational Mathematics & Statistics | 2012-2017
Publications¶
Liu, J., Weitzman, E.R., & Chunara, R. (2017). Assessing behavior stage progression from social media data. CSCW 2017, 1320-1333.
Rehman, N., Liu, J., & Chunara, R. (2016). Propensity score matching for vaccination sentiment analysis. AAAI Spring Symposium.