Experience
Work Experience
-
Working with Professor Shangtong Zhang
to create a solution to a multi-objective sensor usage optimization problem using reinforcement learning
-
The novel aspect of this work is that the agent must learn to balance the tradeoff between
the value of knowing its current location by turning on its GPS and the cost of the GPS sensor draining the phone's battery,
and it must do so without knowing the current location of the phone (it must choose an action without knowing the current state)
-
So far the project is using simulated data but the goal is to apply this to realworld phones and also with multiple instances and types of sensors.
As there are many multi-objective reinforcement learning methods, the goal is to find or create the best one for this problem.
Skills learned: Pytorch, Multi-objective optimization, research and design, reinforcement learning
- Designed a per-data-set discussion page to facilitate collaboration between Capital One associates
- Deployed the page and is being used in over 19,500 data-sets and by over 15,000 monthly users
- Eliminated hundreds of hours of development time for the team by creating a faster local development environment
- Presented design improvements to product designers when oversights were discovered
- Skills learned: Apollo GraphQL, React, Tailwind CSS, Webpack, UI/UX, reviewing pull requests
- Worked as a teaching assistant for undergraduate computer architecture
- Lead 100+ students in laboratory sessions and office hours by conducting code reviews and peer mentoring
- Improved my ability to read code, debug, explain complex concepts, master low level programming skills, and lead classroom sessions
- Designed and deployed an automated server-less system to audit vulnerability report data using Amazon Web Services
- Created a responsive chart user-interface to display reports for internal capital one customers
- Improved real-time database updates by hashing vulnerability data
- Automated database population by creating API’s to call re-population lambda functions
- Skills learned: AWS DynamoDB, Dynamo stream, Lambda, EC2, API/HTTP, Python, Angular, NodeJS
- Organized and hosted cultural and social events for over 400 members
- Captured, edited, and published thousands of photos and videos from events
- Facilitated social and academic connections between members
- Taught a high school student concepts in graph searching and constraint satisfaction problems
- Taught a middle school student Geometry
- Learned people skills with teenagers from varying backgrounds and skillsets
- Worked with Dr. Hongning Wang and PhD student Zhendong Chu on machine learning on noisy data
- Implemented both majority voting and a custom layer in pytorch to resolve discrepencies in data labelling
- Custom layer achieved 14% increase in crowd-sourced data classification compared to the baseline majority voting algorithm
- Gained experience reading and implementing papers, designing experiments, and statistical evaluation and presentation
- Skills learned: Python, PyTorch, Tensorflow, Numpy, Linear Algebra, Statistics, Machine Learning, Data Science