Stanford, California, United States
I'm Felix, a PhD student at MIT working on AI for health. I began my journey with CS as a student at AEOP UNITE, a pre-college research program where I conducted novel research with like-minded students. I served as the program manager the following few years and grew my location to what Department of Defense representatives ranked as one of the best in the country. I'm excited to continue using my experiences to broaden my impact, working with labs to understand how computation can benefit cognitive science and cancer research.
Modeling the collective behavior of heterogeneous scalable unmanned systems.
Currently working with Dr. Eric Stice on: “Determining Eating Disorder Chronicity from Novel Baseline Risk Factors” (Zhan, Stice) “Enhancing Effectiveness of a Translational Neuroscience Obesity Prevention Program.” (Stice et al.) “A Brief Dissonance-Based Type 2 Diabetes Prevention Program for Adults with Prediabetes.” (Stice et al.)
Working with Dr. Hans Breiter on: “Using a 5-minute picture rating task for personalized intervention channel selection.” (Zhan et al.) “Emotion as a social signaling system: working towards a computational theory of mind.” (Zhan et al.) “Predicting externalizing behavior with small sets of interpretable reward behavior and survey variables.” (Bari et al.) “Robust interpolation in cognitive computational artificial intelligence: fixing the problem of small datasets.” (Avant et al.) “Judgment variables strongly differentiate between anxiety and depression.” (Vike et al.)
Designed and developed a full-stack chat application in 3 months (typically a 9+ month, 6+ FTE effort), using React, AWS Bedrock, Timestream, and Lambda, to handle AWS CloudFront Vanguard queries and analyze attack data, reducing potential on-call FTE load and streamlining onboarding time for new SDEs by lowering the barrier to entry to identify and access critical system metrics. Consulted 10 teams across Amazon to gather insights and ensure best practices, delivering thorough documentation and future-proofing the tool for continued use and development.
Constructed a specialized pipeline alternative to Ethovision, the industry standard, for the Danzer Lab, analyzing raw pose estimation data from DeepLabCut, conducting comprehensive user interviews and creating clear and concise documentation to ensure accessibility and encourage interdisciplinary collaboration.