Experience
Berkeley, California, United States
Scribe (https://github.com/callaunchpad/Scribe)
Trained a new model in PyTorch that we call Writer Differentiable Varying Length Handwritten Text Generation, which combines the ScrabbleGAN and DeepWriter models into one overarching GAN architecture jointly trained with Reptile, a meta-learning algorithm, to generalize better and enable few-shot learning of handwriting styles
Strategist (https://github.com/callaunchpad/Strategist)
Developed quantitative trading models using the Backtrader Python library based on mean reversion theory and using technical indicators like Bollinger bands, moving average convergence divergence (MACD) and relative strength index (RSI)
Inspector (https://callaunchpad.github.io/Inspector/)
Published the Google Chrome extension Inspector (callaunchpad.github.io/Inspector/) to allow users to debunk fake news while browsing the web with just a click. Serverless backend is supported by Google Cloud, and the backend NLP models were trained in TensorFlow and used an architecture combining CNNs, LSTMs and Attention Mechanisms
Seattle, Washington, United States
Amazon Elastic Compute Cloud (EC2) / Enterprise Provisioning Solutions (EPS) Organization - Applications Team.
Designed and optimized SQL, NoSQL and graph databases using MySQL, Amazon DynamoDB and Amazon Neptune, respectively, from the ground-up focused on model flexibility and read-efficiency to pool in/combine multiple existing entities of data, e.g. node type, region, AMI, instance type, etc., to recommend available workload configurations for a specific application, e.g. SAP Hana, from all AWS Regions.
With that, I created a micro-service that consists of Lambdas to consume the data lake and an internal-facing API built using API Gateway to allow the Lambdas to be consumed by other AWS Tier 1 services as well, for 'instance' Launch Wizard.
This project will enable greater efficiency within the AWS platform and faster development time within AWS teams working on projects that interact with EC2/EPS.
San Francisco, California, United States
Worked directly with the CEO and CTO (Will and Yassin) as well as the frontend team in coordinating and implementing real-time production-level frontend and backend changes to the app and servers
Added multiple new food delivery services to the Python backend and Firebase database, with the most notable few being BeyondMenu, Chowbus, Starbucks, Popeyes, Instacart, Anycart, and Shipt
Modified an entire server (10000+ lines of code) to be object-oriented as well as asynchronous/non-blocking via Asyncio and concurrent/multithreaded via Multiprocessing as well as implemented Disjoint Set Union/Union-Find in place of a quadratic brute-force algorithm in a core, bottlenecked post-processing function, which reduced overall runtime speeds on average for users empirically by over 5x
Modified another server (5000+ lines of code) to be asynchronous/non-blocking as well
Monitored 4 production Heroku servers, debugged error logs, and pushed up real-time fixes daily
Debugged runtime errors and memory leaks with the help of libraries like Pympler, Objgraph, and Tracemalloc
Berkeley, California, United States
Provided consulting services to Petal, a VISA credit card company backed by WebBank Inc., to better reach their target audience of college students
Analyzed the credit card industry, compiled 50 pages worth of research, delivered multiple summative decks to peer consultants, redeveloped the client’s website as well as created an assistive chatbot alongside it, and helped present our final marketing and product development strategies to the client’s executives and lead engineers
Education
University of California, Berkeley