Calen Walshe
Quantitative Research Leader · AI/ML for Ads Ranking & UX at Meta
About
Calen Walshe is a survey scientist and applied ML researcher who builds end-to-end systems for turning human feedback into production signals. He combines classical survey methodology, causal inference, and NLP to derive insights from open text and implicit response data. At Meta, he leads the design and deployment of scalable pipelines that feed quality metrics and sentiment signals directly into core ranking and recommendation models. He is deeply attuned to measurement reliability, drift detection, and end-to-end observability in real-time serving environments. His technical fluency spans from conceptual survey design to engineering high-performance inference systems, bridging behavioral science and infrastructure.
Selected Publications
Experience
- Designed and ran large-scale A/B and quasi-experiments to evaluate new product features, directly informing multi-billion-dollar revenue decisions.
- Defined and operationalized north-star metrics for ad quality and user experience that became standards across product and engineering teams.
- Built production data pipelines in SQL and Python to integrate surveys with experimentation, enabling self-serve analytics at org scale.
- Partnered with product and engineering on trust and safety systems to balance fraud risk, implementation cost, and customer experience.
- Deployed large language model pipelines to classify millions of pieces of user feedback, creating new quality signals for ranking models.
- Delivered dashboards that merged behavioral data, surveys, and ML outputs, adopted by hundreds of stakeholders for decision-making.
- Served on the Ads core leadership team, shaping strategy and execution for a group responsible for a significant share of company revenue.
- Built software classifiers to automatically detect poor-quality retinal scans, improving diagnostic reliability.
- Collaborated with hardware engineers to embed the detection system into devices, enabling real-time rejection of invalid scans.
- Contributed to patent US20250057414A1 on retinal disease detection methods.
- Designed computational models of human perception using Bayesian inference, machine learning, and large-scale neuroimaging data.
- Built reproducible pipelines and simulation frameworks to study cognitive performance and visual attention under uncertainty.
- Published first-author research in venues such as Current Biology and AAAI, translating findings into perception and decision-making algorithms.
- Partnered across neuroscience, psychology, and computer science to prototype ML-driven methods and secure multidisciplinary funding.
Education
- Ph.D., University of Edinburgh (2015)
- B.A. (Hons) in Cognitive Science, Simon Fraser University (2009)