Jack White standing in a field, with a bridge and hills behind him

Purdue University · CS & AI · 2023–2027

I deploy AI systems for task automation and conduct robotics research in the field of imitation learning.

I also deploy automatic trading bot experiments trading high volatility small caps. I’m also a semiprofessional keyboardist and vocalist with over a hundred shows under my belt. I’m currently in 3 bands.

Selected work

All work

AlgoBuilder

2025–2026

An LLM-based agent system at BPM Microsystems that generates and validates NAND/NOR flash programming algorithms.

Python
Cursor agents
MCP
Salesforce
BPWin REST/WebSocket

ControVirtual

Apr 2026

XR teleoperation and a fine-tuned SmolVLA policy for an SO-101 robot arm, built at StarkHacks 2026.

Unity
Meta XR
Python
OpenCV
LeRobot

TradingSim

2026

A browser-based replay simulator for historical market data, with Level-2 depth and no lookahead.

Python
FastAPI
Databento
Parquet
JavaScript

Experience

BPM Microsystems

Jul–Aug 2025; May–Aug 2026

Automation / AI Software Engineering Intern · Houston, TX

Built AlgoBuilder, an LLM agent system that generates and validates NAND/NOR flash programming algorithms. It reached over 80% fully automated completion across more than 40 validated algorithms, saving an estimated 10–15 engineering hours per algorithm.
Modernized Chip, an internal AI assistant used by about 75% of employees. Connected it to Salesforce and Google Drive through the Model Context Protocol (MCP), added tests, and cut the cost of model calls.
Built PDR Builder, a pipeline that turns datasheets and product specs into engineering artifacts, with deterministic validation and human review. It passed 27 of 27 historical design validations.
Extended the BPWin REST API from 1.2 to 1.9 with job monitoring, device discovery, and validation endpoints so programming jobs can run unattended.

Purdue University, Prof. Yexiang Xue

Present

Robot Learning Research: Zero Shot and Active Learning · West Lafayette, IN

Developing methods that use LLMs and diffusion policies to generate robot demonstrations without human input. The demonstrations bridge transitions between hubs, and from hubs to goals, in latent space so a robot can learn new tasks zero shot.

Purdue University, Prof. Yung-Hsiang Lu

Aug–Dec 2025

Computer Vision Research · West Lafayette, IN

Implemented computer vision methods that transcribe piano performance video into MIDI, tested on the Rach3 and PianoYT datasets and used to evaluate generative systems for music performance.

Music

More work

I deploy AI systems for task automation and conduct robotics research in the field of imitation learning.

Purdue University · B.S. Computer Science and B.S. Artificial Intelligence · 2023–2027

MusicOn repeat

Current favorite songs

Tieduprightnow · Parcels

ResearchPresent

Robot Learning Research: Zero Shot and Active Learning

With Prof. Yexiang Xue. Developing methods that use LLMs and diffusion policies to generate robot demonstrations without human input. The demonstrations bridge transitions between hubs, and from hubs to goals, in latent space so a robot can learn new tasks zero shot.

WorkMay–Aug 2026

AlgoBuilder

An LLM-based agent system at BPM Microsystems that generates and validates NAND/NOR flash programming algorithms.

fully automated completion: 80%+
validated algorithms: 40+
estimated engineering time saved per algorithm: 10–15 h

Work2026

PDR Builder

A pipeline that turns semiconductor datasheets into pin-driver requirement files, with deterministic checks and human review.

historical design validations passed: 27/27
noise errors per design: 0.09
silent errors on scored pin cases: 0

ProjectApr 2026

ControVirtual

XR teleoperation and a fine-tuned SmolVLA policy for an SO-101 robot arm, built at StarkHacks 2026.

teleoperated episodes: 50
task success after fine-tuning: 0 → 75%
Best Use of AMD Tech, honorable mention: AMD

Opened for Magdalena Bay

With 3rd Street Collective at Purdue’s Elliott Hall of Music, for about 6,000 people.

Project2026

TradingSim

A browser-based replay simulator for historical market data, with Level-2 depth and no lookahead.

automated tests: 126
Level-2 depth: MBP-10

RCBot

A small-cap momentum trading bot that runs the same code against historical replay and Interactive Brokers.

event format for replay and live: 1
paper trading: IBKR

Project2026

FilingEdge

A machine-learning research system that tests whether SEC filings predict abnormal returns, using only data available at filing time.

point-in-time features: 34
leakage and cost checks: pytest

SlopMop

A browser extension that flags likely AI-generated posts on LinkedIn, Facebook, Instagram, X, and Reddit.

social platforms: 5
automatic detection time: ~0.6 s

Recipe Finder

Upload PDF, EPUB, or TXT cookbooks, extract the recipes, and search them by what is in your pantry.

cookbook formats: PDF, EPUB, TXT
rotating refresh cookies: JWT

ResearchAug–Dec 2025

Piano transcription

Computer-vision transcription of piano video to MIDI, used to evaluate AI-generated performance videos.

datasets: Rach3, PianoYT
Purdue research conference: Fall 2025

ProjectOct–Dec 2024

C Shell

A Unix-style shell written from scratch in C with Lex and Yacc.

lexer, parser, and execution: C

TeachingApr–Jun 2023

Music Teacher, Redwood High School

Taught music theory, ear training, and piano technique across classical, rock, and jazz.