Vijay's Personal Website
My Projects
| Name | What is it? | Link |
|---|---|---|
| RISC-V CPU | Built a full CPU in Logism with two-stage pipelining | View |
| Handwritten Digit Classifier | ML algorithm in RISC-V assembly to classify digits | View |
| FactGrid Cuneiform Project | ML model to segment and transcribe Cuneiform tablets | View |
| The David Deutsch Dictionary | Online database of quotes with Google Sheets backend | View |
| Scheme Interpreter | Python-based evaluator and interpreter for Scheme | View |
| Helium Balloon Project | OpenCV-powered object tracking for helium balloons | View |
| BYOW (Build Your Own World) | Pseudo-random world generation algorithm in Java | View |
| 2D Convolutions Optimizer | Parallelizable 2D discrete convolutions algorithm | View |
More Info
| Project Name | Image | Description |
|---|---|---|
| RISC-V CPU |
|
Built a full CPU in Logism that runs instructions in assembly language. Implemented two-stage pipelining to increase throughput. Prevent control hazards via pipeline flushing. |
| Handwritten Digit Classifier |
|
Designed machine learning algorithm to classify handwritten digits from MNIST database of 60,000 28x28 images. Wrote RISC-V assembly procedures for matrix multiplication, dot products, and the activation and loss functions. |
| FactGrid Cuneiform Project |
|
Develop machine learning model to segment and transcribe dataset of 1,068 annotated Cuneiform tablets using the YOLOv8 architecture. |
| The David Deutsch Dictionary |
|
Online database of quotes from prolific science author and quantum physicist David Deutsch. Utilizes Google Sheets as a backend database to allow for seamless and near-real-time updates. Scraped quotes off the internet using Python3 and NLTK (Natural Language Tool Kit). Designed UI using Flask and HTML. |
| Scheme Interpreter |
|
Python-based evaluator and interpreter for Scheme, a dialect of Lisp. |
| Helium Balloon Project |
|
OpenCV-powered object tracking for helium balloons using the CSRT (Channel Spatial and Reliability Tracking) algorithm. |
| BYOW (Build Your Own World) | Pseudo-random world generation algorithm built in Java. Also created a UI for generating, saving, and loading worlds. | |
| 2D Convolutions Optimizer |
|
Created algorithm for running 2D convolutions. Sped up algorithm by a factor of 8.10x using SIMD Instructions, multithreading via OpenMP, and targeted algorithmic optimizations. Wrote OpenMPI coordinator to support process-level parallelism capabilities. |
Bio
My name is Vijay (“VJ”) Kethanaboyina, and I’m a computer science student at UC Berkeley aspiring to work as a software or data engineer. I’ve solved problems in technical areas such as computer vision and Linux system administration. I’m also an avid reader and language learner.
Links
- Reading Log
- RSS (Atom) Feed: vkethana.com/feed.xml
- GitHub: github.com/vkethana
- Email: vijaykethanaboyina at gmail dot com
Education
- Computer Science BA from UC Berkeley (class of 2026).
- Computer Programming Associate’s Degree from Gavilan College (concurrent enrollment in high school)
Skills
- Languages: Python (Pandas, Matplotlib, Plotly, OpenCV), C, Assembly, C++, Java, R, HTML/CSS, JavaScript, SQL
- Frameworks: React, Node.js, Flask, JUnit
- System Adminstration: Debian, Arch Linux, Vim, DigitalOcean
Experience
Undergraduate Researcher for UC Berkeley Sky Computing Lab
June 2024 - Present
- Collaborating with UC Berkeley CS PhD students to enhance R2E, a framework for evaluating code generation capabilities of large language models. Scaling Python code generation benchmarks to include snippets from a wide selection of repositories (end goal: 1000+).
- Developing an LLM-powered agent to automate the cloning, setup, and execution of Python GitHub repositories; detecting and fixing installation errors using GPT-4 powered agentic loop.
Undergraduate Researcher for FactGrid Cuneiform Project, UC Berkeley
August 2023 - February 2024
- Developed machine learning model to segment and transcribe Old Assyrian Cuneiform tablets using the YOLOv8 architecture, under the mentorship of Dr. Adam Anderson (link to project poster)
- Designed pipeline to convert nearly 3,000 raster images of Cuneiform signs into vector images using PIL, Inkscape, and Potrace, and then uploaded them onto a publicly-accessible database using MediaWiki API.
- Created OCR model to support development of language models for under-resourced languages written in Cuneiform, especially Old Assyrian. Final product used to annotate dataset of 1,068 tablet images.
Hackathon Cofounder
2022-2023
- Organized Gilroy Hacks, a multi-school, student-led hackathon series for middle/high schoolers in Gilroy, CA.
- Secured a $1,481 grant by convincing sponsors and resolving logistical issues. Coordinated with judges and backers, including Wolfram Language, 1Password, and AoPS.
- Support development of 7 hackathon projects in Python, Next.js, Typescript, and Flask.
Future Projects
- Continue with my CS coursework (I plan to take machine learning, convex optimization, probability and random processes, and operating systems next year), with some Sanskrit on the side.
- Write more stuff on my blog.