Palo Alto, California, United States
Research-oriented ML engineer specializing in large-scale sequence modeling, retrieval systems, and evaluation design. I've spent the last few years finding the gaps between how models train and how they actually behave in production, while building the research directions to close them. Motivated by AI systems that are both capable and reliably aligned with intended behavior. https://www.mlawrence.info
YouTube Shorts, Recommendations TL for flagship retrieval model. Responsible for end-to-end model quality, including model architecture, data, and training/serving infrastructure. * Led long-term initiatives for retrieval model quality improvement, leading to substantial improvement in YouTube Shorts engagement. * Worked horizontally to improve modeling and infra for other models in YouTube Shorts, finding many training/serving improvements for other key models in the ecosystem * Mentored many other machine learning engineers
AutoML/Vertex AI MLE on model platform team. Converted numerous in-house models into new Cloud products as part of a research partnership. Involved in deep learning for tabular data, time series forecasting, and eventually large multimodal models.
Machine learning engineer on biotech demand forecasting platform.
- Invented novel statistical method for identifying hold messages and scripted language [Patent pending] - Created safety tooling to check thousands of live machine learning models before a deploy - Automated model health checks for product managers with a PDF of performance metrics - Led a team of talented undergrad data scientists for UCSB’s data science capstone initiative
Tutoring for math, physics, and chemistry from high school to university-level coursework.