![video_snapcut](images/video_snapcut.jpeg)
Video Snapcut
Computer vision video cutout tool based on Video SnapCut algorithm from Bai, Xue, et al. paper. A MatLab pipeline for propogating object segementation from a single frame throughout a video.
About Me
Bio I used to know math, and I liked that. Know I go on long walks. My life goal is to Discuss Great Things.
Experience Google
College University of Maryland, College Park
Major B.S. Mathematics and B.S. Computer Science
Age 22
Google SWE II: Revamped data collection and cleaning collection. Launched new production logging system for real time metric collection on platform usage, leading to data analytics over user sessions. Onboarded new user to our data catalog, overseeing database migrations and updates. Launched new monitoring process for automatic alert generation of assets not meeting their SLO.
Perceptual Interfaces Lab: Research on using neural networks for approximations of theoretical signal reconstruction techniques. Developed signal analysis specific network architectures using PyTorch for simulation and training.
Google SWE Internship: Migrated distributed catalog system to new server platform. With this, and input from stakeholders, designed a new developer portal for learning about recommended best development practices.
Institue for Systems Research: Investigation of machine learning algorithms as metareasoning in autonomous multi-agent systems. Implemented distributed task allocation algorithms for new multi-agent optimization scenario and ran as a event-based simulation to generate data for training the metareasoning model.
Google Step Internship: Launched new feature for heuristic monitoring of production logs, tripling services coverage. A MapReduce pipeline using statistical analysis to create data invariants over noisy numeric distributions.
Balloon Club: The UMD Payload Program develops scientific payloads to be flown on Weather Balloons in nearspace. My payload team, SATRN, designed a payload for the remote and reliable release of other payloads on the weather balloons tow.
Rocket Club: The UMD Rocket Club tests and designs rockets that fly in the 10,000 to 30,000 feet range in the atmosphere. Worked on the interface for the flight data collection.
Strengths:
Languages: Python, Java, C, SQL, OCaml, MatLab, JSON, Kotlin, Bash, Typescript, Jinja, html, CSS, Markdown
Systems: REST, RPC, Protocol Buffers, MapReduce, Relational Databases, Vector Databases, PyTorch, Docker, Dependency Injection, Numpy, OpenCV, ROS, Arduino, React, Angular, Flask, Bazel
Some of my latest personal projects:
Computer vision video cutout tool based on Video SnapCut algorithm from Bai, Xue, et al. paper. A MatLab pipeline for propogating object segementation from a single frame throughout a video.
Web app for keeping up to date with their favorite bands. Users can follow a band a to receive email updates whenever new music is released.
Collaborated with client, the Bay-Ridge Senior Center, to create an AWS hosted web application for unifying their data collection with automated communications.
A CLI to send torrents over gmail, allowing clients to automatically begin download process of very large files. Not the most useful development, but good experience with torrents and MAC OS launchd service management daemon.