📝 🙋🏻♂️More about me.
I have over 8 years of experience in data science and machine learning, working with Stitchfix, Facebook, Kojo, NYU, ActionIQ, Meltwater and multiple startups. I love learning new technologies and just making stuff, but at work I like to stay pragmatic. In 2023, I have been learning JavaScript to build ML products for engaging user experiences in the browser and I like to stay full stack. Check out some demos I've been working on.
South Park Commons - Exploring LLMs — 2023
- Exploring applications and techniques for using LLMs in production settings
- Launched Youtube University, a better way to do long video summarization to useful notes, creates charable chapters, using timestamps and markdown, over 6k unique visits a month.
- Developing Journal a daily journaling app powered by a AI life coach with long term memory.
Stitchfix — Staff ML Engineer — 2018-2023
- Pioneered the use of OpenAI’s LLMs to increase the efficiency and productivity of human stylists with prompt engineering. Streamlined the writing process for stylists, improving turnaround time and writing and recommendation quality.
- Lead computer vision research at Stitchfix, managing a labeling task force, fine-tuning vision models (bounding box and search), and writing crawlers to create an initial dataset of over 300,000 outfits to QA which ultimately resulted in the outfit dataset for Stitch Fix.
- Developed a vector-based product similarity search system using CLIP, GPT-3 embeddings, and FIASS. This system significantly improved the product carousel recommendations and led to over 50 million dollars in annual revenue.
- Architected Flight, a recommendation on rails framework which improved reliability, maintainability, response times, and scalability with over 80% adoption in across the org and services 350 million requests everyday. Learn more
- Established deployment best practices, prototyped new products, and provided technical guidance during the company's shift to a real-time, direct-to-customer model.
Facebook — Data Scientist: Protect and Care — 2017
- Developed algorithms to identify and rank potentially harmful content on the platform, enabling human moderators to ensure a safe user experience, mostly focused on violence, hate speech and child endangerment.
- Built dashboards to uncover coordinated malicious activities aimed at important accounts and viral content in order to minimize the PR Risk of Facebook.
Education
Sept. 2012 – Dec 2017 B.MATH. COMPUTATIONAL MATHEMATICS AND STATISTICS University of Waterloo
Waterloo, Ontario, Canada.
Conference Proceedings
- 1 Liu, J., Weitzman, E. R. & Chunara, R. (2017). Assessing behavior stage progression from social media data. In Proceedings of the 2017 acm conference on computer supported cooperative work and social computing (pp. 1320–1333). ACM.
- 2 Rehman, N., Liu, J. & Chunara, R. (2016). Using propensity score matching to understand the relationship between online health information sources and vaccination sentiment. In Proceedings of association for the advancement of artificial intelligence spring symposia (pp. 23–25).
Jason Liu Resume (1)