Took a professional risk joining a small seed-stage startup as engineer #3. Harvested Financial was the first robo-advisor to manage financial derivatives for retail investors. Built core infrastructure, engineering processes, and data security practices.
designed and implemented a security strategy for encryption of all customer PII including SSN, address, tax filings, DOB, etc using client-side field level encryption
developed re-usable event-driven architectures using GCP Cloud Scheduler, GCP Pub / Sub, and Compute Engine for ETLs that batch process customer data
developed core CI / CD pipelines using GCP Cloud Build / Bitbucket and latest python testing frameworks (tox, pipenv, pytest, etc)
I learned a lot about financial derivatives, options markets, early-stage startup business needs, and building financial applications.
Developed the machine learning platform at 23andMe, which uses ML to provide predictions to customers on their risk for genetically-linked diseases. Built tools to speed up the iteration cycle for engineers and data scientists and solutions for the end-to-end ML lifecycle: training, deploying, bulk computing, serving, monitoring, and model management.
developed automated pipelines and datastores that monitor the performance of models in production and display statistical metrics in both custom web applications and Databricks MLFlow
scaled services for training & serving models on large genomic datasets using Elastic Container Service (ECS) and AWS Batch
built a real-time prediction service with 200ms latency for model serving while increasing the infrastructure limit for both larger and more accurate models by 10x
I learned how to build and architect machine learning applications in AWS and tools for data scientists
San Francisco Bay Area
I worked for the data science team at Socos, a startup that combines behavioral psychology and machine learning to make recommendations for parenting and child development.
modeled the latent factors for neighborhoods (zipcode specific) using public information for IRS tax returns and demographic data
developed the personalized recommender system for survey style questions that users receive everyday
clustered similar user profiles, imputed answers to questions users haven’t answered, and leveraged the semantic relatedness of questions
created visualizations for user engagement and profiles on personality attributes of children
I learned how to take ML into production and what’s state-of-the-art through reading papers and great mentorship.
Foster City
I worked for Visa Token Services – the API for Apple Pay and Google Wallet. I prototyped a dashboard for my manager to compile # of open bugs, release stages, health of different environments, etc., across teams in one place. Using a MEAN stack, I mostly developed the front-end views & charts using Angular, but I also contributed to the Mongo ‘schema’ (although Mongo is schema-less) and created new APIs for additional features. I learned a lot about security in the payments industry.
Education
University of California, Berkeley