Work
You can find my up-to-date work experience on my LinkedIn.
Software Engineer
Mountain View, CA
Jun 2024 - Sep 2024
- Engineered high-quality datasets of 50+ examples for clinical tasks such as medical condition summarization and conversational diagnostics (AMIE) using advanced data cleaning techniques and clinician-in-the-loop feedback.
- Fine-tuned medically specialized LLMs using the curated datasets, enhancing model performance through iterative optimization and validation.
- Developed an auto-evaluation system leveraging prompting techniques like chain-of-thought, structured decoding, and critic agents that achieved a 0.8 correlation with human evaluations, streamlining the assessment of fine-tuned LLMs for accuracy, coverage, and clinical safety.
Product Manager
Mountain View, CA
Aug 2022 - March 2023
- Led development of a new clinical risk scores feature that enables physicians to automatically compute scores in the context of relevant patient data.
- Authored a product requirements document (PRD) that defined the scope, objectives, and key performance indicators (KPIs) of the feature.
- Conducted and analyzed user research (UXR) to understand the needs and pain points of physicians regarding risk assessment.
- Presented the feature design and roadmap to the Director of PM and VP of Google Health. Received positive feedback and approval to proceed.
Software Engineer
Mountain View, CA
May 2022 - Aug 2022
- Designed, implemented, and launched an automatic pipeline that enriches clinical annotations with metadata from medical ontologies (e.g. SNOMED, UMLS), reducing end-to-end time to launch conditions in product by 50%..
- Employed medical expertise to curate analytics for 12 chronic conditions, successfully launched within the Care Studio product.
- Developed an AI model leveraging medical ontology data to forecast relationships among diseases, medications, labs, and procedures.
Curai Health
Machine Learning Engineering Intern
Remote
Aug 2021 - Dec 2021
- Collaborated with clinicians and engineers to develop novel, clinically-informed metric for evaluation of an internal entity linking (EL) model.
- Proposed and implemented production-level changes to the EL model, including an ML-based medical spell checker. Validated these changes using the new metric.
- Shared results with the ML team, Clinical Innovation team, and CTO. Documented detailed next steps for future model versions.
Mass General Brigham
Research Assistant
Boston, MA
Jun 2021 - Aug 2021
- Trained novel medical image segmentation models using PyTorch to isolate kidneys and ureters on CT scans and grade lumbar stenosis on MRI scans.
- Presented findings to stakeholders from GE Healthcare and MGH and packaged models into deployable microservices using Docker.
epistemic.ai
Machine Learning Engineering Intern
Remote
May 2020 - Aug 2020
- Trained a Graph Neural Network (GNN) to generate biomedical knowledge graph embeddings, optimizing downstream information retrieval tasks.
- Delivered proposal and proof-of-concept to the CEO and laid the foundation for integration into the product's retrieval service.
Brown AI Lab
Research Assistant
Providence, RI
Dec 2018 - Current
- Built de-identification system using TensorFlow/Keras with 94% accuracy at identifying HIPAA-regulated tokens in Electronic Health Records. First-author paper accepted in American Medical Informatics Association (AMIA) Informatics Summit.
- Developed time-series AI model to predict and forecast delirium phenotypes in stroke patients using time-series data collected from wrist actigraphs. First-author paper accepted in Frontiers in Neurology.
- Led six-person team to build a search engine for clinical trials. Presented work at Text REtrieval Conference (TREC) Precision Medicine Track.