Researched, designed, and implemented a deep reinforcement learning agent which can interact with any linux-based computer application and circumvent commercial anti-cheat gaming software. The Agent consists of several concurrent unix processes continually leveraging OpenCV for computer vision, Pytorch for deep learning, and Linux drivers for simulating device input. The agent is currently able to increase player engagement metrics by 3x.

Performed as the primary contributor to a C++ and CUDA project dedicated towards exploring the efficacy of neuroplasticity in neural networks. Highlights include:- Implementing basic feedforward neural network architecture in CUDA to allow for efficient weight transfer between differing network architectures- Leveraging classical AI algorithms to prune weak connections between neurons and introduce new neurons into the network architecture

I performed as a backend engineer developing mostly in Python. I was primarily responsible for end to end ownership of new API features and automating developer workflows. Some highlights include:- Saving one week per month in critical payment releases by introducing automated data quality checks for contract data that was previously manually tested.- Improving client experience by creating a new report providing key data insights regarding why payments changed with finer detail about membership, risk, and quality than was previously available.- Leading redesign of a payment engine to assess providers from a common set of metrics, implement said scoring system to be less exploitable, and standardized its measurement data schema.

- Designed and implemented the user subscription cancellation experience for subscription boxes- Lead the design process with an emphasis on low latency and extensibility- Performed as a full stack engineer coding Java controllers for the backend and TypeScript, HTML, and CSS components for the frontend

- Provided business strategies to help improve the win rate of construction contracts- Strategies were based on analyses of historical contract data using numerous machine learning algorithms ago predict outcomes and explicitly weight characterizing variables

- Wrote production-level AWS Lambda code in Python to make daily updates to an AWS Redshift database via AWS S3, Parameter Store, and Marketo's API- Integrated the code to Everspring's production infrastructure using Terraform

- Selected from a pool of thousands of candidates to work closely with the Jumpstart team to build out a machine learning recruiting platform - Pilot new features, provide student feedback and act as a product advisor on campus - Leverage various growth strategies and tools including email marketing, social media campaigns, and presentations/demos - Network with tech companies to bring industry professionals and recruiters on campus - Help to coordinate logistics for campus events and provide peer mentorship

Reviewed and improved upon an existing system to keep track of customers' data and communicate with client servers

Worked with Canvas LMS's API and AWS Buckets to write an algorithm to periodically backup course data