Hi, I'm Brandon!
I'm a senior in Columbia University, studying computer science and entrepreneurship. My background consists of AI research, corporate experience, and tech startups. If you're interested, please contact me here.
I'm a senior in Columbia University, studying computer science and entrepreneurship. My background consists of AI research, corporate experience, and tech startups. If you're interested, please contact me here.
Fine-tuned and evaluated two 8B-parameter LLMs on 20k synthetic documents, measuring a 0% → 5% increase in unfaithful chain-of-thought behavior via controlled evidence-citation tasks and identifying accuracy trade-offs for sensitive document workflows. Paper
Skills learned: Finetuning, research analysis
In the A2R lab, I made and debugged firmware in C++ and C for the Crazyflie drone, optimizing tinyMPC for model predictive control. Resolved memory constraints, build issues, and ensured integration with Crazyswarm2 (ROS 2)
Skills learned: Embedded systems, firmware
Designed and deployed a scalable schema for crackd, supporting dorm and university data beyond Columbia's campus. Used Supabase, PostgreSQL, and TypeScript; also built evaluation tests and infrastructure to analyze LLM understanding of humor and cultural context
Skills learned: Typescript, LLM eval
Built dynamic Excel models and automated workflows to track real-time financial progress for a large industrial goods client; presented quantified results to BCG and client leaders, enabling informed decision-making
Skills learned: Automation, client communication
As a TA for the User Interface (UI) Design course, I was responsible for conducting weekly office hours, evaluating assignments, and answering student questions on class discussion board
Skills learned: Mentorship, communication
At FluidityIQ, I evaluated vectorizers for patent claims, built the patent dataset, enabled multi‑user chat API support, and shipped to production - cutting inference latency 6x. I also prototyped an AI agent using Milvus and Haystack
Skills learned: Langchain, Azure
I debugged entitlement and data-access APIs for a U.S. bank; improved web features using REST and SQL, fixed permission bugs, and automated test-data creation in Python. Presented outcomes to 30+ executives.
Skills learned: Python, Java
Implemented 4 ML models to predict the lifespan of electronic parts and presented the results to 20+ executives. Also developed 8 ML models with AWS to forecast raw material prices. Codebase
Skills learned: ML, AWS
I worked on an image processing project for Artaic's mosaics, training a YoloV5 model with 900 tile images, achieving 100% accuracy at 90% confidence. Codebase
Skills learned: Computer vision
Co-founded a hands-on LLM builder community in collaboration with Anthropic; organized technical workshops on prompt design, evaluation, and model experimentation
Skills learned: MCP, Community-building
Made a browser extension that provides hints for Leetcode problems. Reached 200+ users! If you're interested, please try it here
Skills learned: Full stack dev, MongoDB
In the Almaworks Accelerator, we've paired 30+ Columbia student founders with industry mentors & VCs. We've hosted five Demo Day events, enabling the startups to present and network.
Skills learned: Presentation, entrepreneurship
Worked in the Pacbot team for Columbia Robotics,
fine-tuned the A* pathfinding algorithm and implemented Q-learning for navigation
Skills learned: Control algorithms, reinforcement learning
