Deep Neural Networks, CNN, GAN, LSTM, NMT, Transformers/BERT.
DQN, REINFORCE Policy Gradient, Actor-Critic, DDPG.
React, Jest, iOS Swift, Android.
Python, SQL/RDBMS, AWS, Flask/Node.js+Express., FastAPI, CircleCI
Snorkel AI
Jun, 2022 — Sep, 2022
- Unicorn-funded startup, focusing on data-centric AI through foundation models and weak supervision
- Enabled ground truth versioning, allowing users to revert back to working states and recreate previous models
- Implemented Application DAG versioning, allowing users to manage previously created DAG structures
- Developed a variety of improvements for unsolved user problems such as proper error surfacing and transparency.
Snorkel AI
Jun, 2021 — Sep, 2021
- Built a multitude of core functionality for Snorkel Flow platform that allows label generation via weak supervision
- Enabled trainable preprocessors, to allow operators to be fitted to application datasets.
- Implemented user-defined operator classes, allowing clients to export their own operators via application DAG
- Implemented user-defined model classes, allowing registration of custom experimental models via Jupyter Notebook
- Developed a variety of improvements for unsolved user problems such as estimated model training time and operator namespace registration conflicts.
Ezra
Jun, 2020 — Aug, 2020
Ezra is an AI-powered financial advisory mobile application that scrapes SEC, twitter, and news data to provide long term investment recommendations. link
- Implemented and trained graph recurrent neural network on S&P 500 price and sentiment data using PyTorch.
- Implemented DCF Model to forecast intrinsic value of portfolio and predictive sharpe ratio optimization model (Backtests performed 15% better than S&P 500 on 2018-2020 backtest)
- Built automated tax-loss harvesting algorithm to save user's money on income and investment gains tax.
- 200+ users reserved for private beta launch.
VMware
Jul, 2019 — Jun, 2020
Engineered several Supervised Machine Learning models and Reinforcement Learning agents to solve problems ranging from automatic security triaging, VM timeout prevention, and VM Anomaly Detection.
- Built Machine Learning models to automate security triaging process of virtual machines and classify true security issues.
- Built Machine Learning models to predict workload timeouts on VMs, enabling proactive recovery actions to be taken.
- Constructed multi-dimensional time-series prediction models to detect anomalies within machine health that would result in high cost and usage
- Developed a suite of Reinforcement Learning frameworks (DQN, DDPG, Reinforce, Actor Critic)
- Designed Deep Learning and PyTorch workshop for VMware community.
- Developed advanced NLP techniques for a lab in an internal ML conference.
twyne
Oct, 2019 — Jun, 2020
Twyne helps users interact with the world through simple and natural motions by bringing automated and customizable ML-based gesture recognition to the smartwatch.
- Developed data engineering (framing and labeling) scripts to process gesture motion data from smartwatch wearable.
- Implemented and trained Spectrogram Fourier Transform of time series data and LSTM model using PyTorch (~ 87 % accuracy)
- Constructed prototype that uses "swipe next" & "swipe back" gestures from smartwatch to navigate through google slides presentation.
- Implementing Siamese metric learning model for per-user gesture customization.
9Eighteen
August, 2018 — June, 2019
9Eighteen is a mobile platform that provides fast, convenient service on golf courses. With one tap, users can order directly on the mobile app and will be promptly service by the closest cart driver. As a developer, I built the full stack of the mobile platform, which is now being used by upwards of 9 golf courses.
Hack Club @ VCHS
Jun, 2017 — Jun, 2019
Worked closely with students to explore the field of computer science and guided them to build their own projects and applications.
Eemergency
Dec, 2017 — Aug, 2018
Eemergency is a mobile platform that tracks live location updates of on-duty emergency vehicles and alerts drivers to pull over to safety. Developed core application; filed three provisional patents on emergency repsonse system architecture.
Carnegie Mellon University
Jun, 2017 — Aug, 2017
Built two games with a team of five: Perspective and DinkTheBlackSun (listed in the portfolio below). Constructed a production level game with Unity .NET
Los Angeles, CA
Sep, 2019 — Jun, 2023
Computer Science Major. Statistics Minor.
San Jose, CA
Aug, 2015 — Jun, 2019
Applied Math Science & Engineering Major.
Los Angeles, CA
Feb, 2019
Built song-guessing host-based game.
Tech Stack: iOS App, Spotify Web API + React Native, SQL + Flask Backend
San Jose, CA
May, 2018
2nd Place Team Award in Math League Nationals.
San Jose, CA
Mar, 2018
Scalable Real Time in Memory Storage and Indexing for Emergency Response System
Dynamic Geofence Algorithm for Emergency Response System
Smart Dynamic Alerting System