Tech Lead Manager at Contextual AI, leading research in LMs, agentic RAG, synthetic data, evaluation, and alignment. Previously founding member of the Alexa LLM post-training team, which became Amazon AGI (led early work on SFT, data collection, and owned RM/eval initiatives). PhD from CMU LTI (advised by Dr. Maxine Eskenazi) working on dialog systems research [thesis] (2022) and BSc from UBC (2018).
As a researcher, I most enjoy doing product-driven research -- I strive to do work that enables next-generation user experiences through strong problem formulation (e.g., schemas in my thesis). Key areas include:
- Systems over Models: Multi-step pipelines [structured fusion '19], agentic workflows, end-to-end optimization
- Evaluation: Natural language testing [lmunit '23], automatic metrics [usr '20, fed '20], realistic user experience evaluation [nsf report '22, interactive eval track '22, dialport]
- Human Control & Specification: Structured representations [schema-guided '21], example-driven approaches [example-driven '21]
- Post-training & Alignment: LLM specialization [apo '23]
- Synthetic Data Generation: Language models as data [lad '22]
Previously held internships at Amazon (2018, 2020; improved intent prediction [example-driven '21] and introduced dialoglue), Meta (2016, 2017; shipped first subword neural MT), and Microsoft (2017). Also spent 2 years as a part-time research assistant in bioinformatics at BC Children's Hospital (graph-based genome representation).
At Contextual, we're building the next generation of LM-powered systems. Always looking to meet exceptional researchers and interns interested in doing ambitious research and making product impact - check out our open roles.