A2R Lab
Worked on Crazyflie drone firmware in C/C++, optimized tinyMPC for model predictive control, and integrated the system with Crazyswarm2 on ROS 2.
RoboticsC/C++
About
My background is in building software for hardware and robotics contexts - control theory (MPC, LQR) in Drake, real drone firmware at Columbia's A2R Lab, XR applications, and low-level systems in C.
I also love building and shipping products. I created Dexter, an AI redaction tool that advanced into sales conversations with BCG and OC&C, and have shipped products like Unscripted and BeetCode reaching 300+ active users.
Robotics & Systems
Firmware and control, machine vision, autonomy, low-level systems.
Cart-Pole Control
Built swing-up and balancing control for an underactuated cart-pole in Drake, comparing MPC against LQR under real actuator limits; ported the MPC loop to C++ for an 8-12x latency cut.
DrakeMPC
Artaic
Built a YOLOv5 model with Roboflow to classify mosaic tile types at 91% accuracy, automating the manual inspection every assembled mosaic went through before shipping.
Computer visionYOLOv5
PacBot 2024
Tuned A* heuristics and built ghost-prediction for Columbia's entry in the 2024 PacBot competition (robot that plays Pac-Man autonomously on a physical maze). Also prototyped a deep Q-learning policy.
PathfindingDeep Q-learning
Unity XR Development
Built a tilt-controlled iPhone maze, a linked AR/VR dungeon exporting scene data from an iPhone builder to Quest 3, and a voice-controlled XR maze using LLM extraction.
UnityC#
Operating Systems
Built a process tracker, custom scheduler, and a file system in C, down at the kernel level. Team project with Connor Espenshade.
CKernel
Products
Built, shipped, and put in front of real users.
Unscripted
Building an AI career coaching platform to help people network more effectively.
AICareer coaching
Dexter
Built an AI redaction tool for consulting slide decks. Piloted with BCG and OC&C, and refined across 15+ consultant interviews to fit real workflows.
AIConsulting
Beetcode
Built and shipped a browser extension that gives LeetCode hints, reaching 100+ users organically.
Browser extensionLeetCode
AI Safety
Tested whether synthetic document fine-tuning reduces chain of thought honesty in Qwen3 and DeepSeek. Both models lost task accuracy and slipped from 100% to 95% faithfulness. Selected for the BlueDot AI Safety cohort.
PyTorchEvaluation
Experience
Software engineering, then enterprise technology consulting.
2021 – 2022
Arrow Electronics — Machine Learning Intern
Designed and ran 200+ ML experiments (Optuna, ClearML) to forecast electronics lifecycle metrics; improved MAPE 68% relative to baseline models.
Enterprise
2023
Capco — Software Engineering Intern
Fixed 66% of open bugs in Java for a major U.S. bank's entitlements service; work integrated into client systems.
Mid-size
2024
FluidityIQ — Software Engineering Intern
Deployed a patent-search vectorizer via Azure Kubernetes CI/CD, cutting API latency 6x. Built an LLM patent summarizer with modular model-swapping.
Startup
2025 —
Boston Consulting Group — Technology & Digital
Automated financial tracking for EBITDA/OWC for procurement, supply chain & operations for an industrial client; drafted an AI implementation proposal for factory operations, presented to the client's CFO and CTO; received fast-track return offer to specialize in tech/digital.
Consulting
Education
Columbia, and what I did outside of class while I was there.
Columbia University
B.S. Computer Science, cum laude — Minor in Entrepreneurship
3.99 GPA · 2022 – 2026
Operating Systems · Advanced Programming in C · Parallel Optimization for Robotics · VR/AR in Unity · Artificial Intelligence · Natural Language Processing
Activities & Leadership
Columbia Spectator
Head of ProductLed a cross-functional product team across three of Spectator's digital products - CULPA, theSHAFT, and the mobile app - running user interviews, defining features, and shipping with engineering.
Almaworks / CORE
DirectorMentored 30+ early-stage Columbia founders - connecting them with the right alumni mentors and investors, and helping them sharpen their pitches across five Demo Days.
Columbia CS Department
Teaching AssistantMentored 25 students a semester building full-stack apps in Flask and jQuery, and introduced a Figma-first workflow that cut development time 50%. Supported 600+ students through office hours and grading.
Claude Builder Club
AmbassadorCo-founded a hands-on LLM builder community with Anthropic and ran workshops on prompt design, evaluation, and model experimentation - teaching peers to build with AI.