San Francisco
I’m a software engineer at Google, improving ads recommendation models.
My broader interest is in safety and fairness in large-scale ML systems. Previously, I contributed to Microsoft Fairlearn and the Seldonian framework, and maintained the COVID-19 Forecast Hub under Prof. Nicholas Reich. It was the official data source for the CDC’s COVID-19 forecasting page.
Before Google I built ML workflows on AWS SageMaker as an applied data scientist at Sway AI, and worked on QuickBooks Mobile as a software engineer at Intuit.
Outside work I hike, run, and play the drums.
A Python library for training ML models and RL agents under the Seldonian framework, as described in Thomas et al., Science 2019. Supports any optimizer from scipy.optimize and ships a CMA-ES implementation.
GoogleSoftware Engineer (ML)
Improving ads recommendation models.
Sway AIApplied Data Scientist
Built end-to-end ML workflows on AWS SageMaker, spanning data preprocessing, training, and deployment. Translated client-specific solutions into reusable components for a no-code self-service ML platform.
Reich Lab, UMass AmherstResearch Engineer
Owned the data validation, visualization, and build pipeline for the COVID-19 Forecast Hub, the official data source for the CDC’s COVID-19 forecasting page. Rebuilt the submission validation pipeline, taking validation time from about 20 minutes down to under a minute and unblocking weekly submissions from 50+ research teams.
IntuitSoftware Development Engineer 2
Shipped on-device ML receipt validation using MLKit, eliminating manual transcription overhead. Built a cross-platform “build once, deploy everywhere” framework adopted by all Intuit product teams, and delivered features end-to-end for QuickBooks Mobile.
University of Massachusetts, AmherstM.S. Computer Science
Coursework: Machine Learning, Neural Networks, Probabilistic Graphical Modeling, Advanced NLP, Reinforcement Learning, Algorithmic Fairness and Strategic Behavior, Advanced Algorithms.
PES Institute of TechnologyB.E. Computer Science
Coursework: Data Mining, Natural Language Processing, Cloud Computing and Big Data.
Ported parts of the CMUSphinx speech-recognition library to Android. Maps speech to identifiers that can be bound to specific actions or triggers. Language-agnostic.
Unified Heterogeneous Hybrid Cloud Management
A vSphere plugin that lets a single interface deploy and manage instances across both vSphere and OpenStack, with a seamless private network between tenant VMs on either cloud.
Caching for Distributed and Hybrid Clouds
Proposes Simple Cloud Federation (SCF), a hybrid-cloud architecture that leverages the hierarchical organization already found in existing cloud systems. Includes an OpenStack implementation.