Edmund

Robots that work in the real world.

ECE student working on solving general-purpose robotics at Princeton.

Who I am

I'm an Electrical and Computer Engineering student at Princeton, minoring in Statistics and Machine Learning and in Robotics, with a focus on the systems that sit between raw sensor data and the decisions a robot has to make: perception, autonomy, and everything in between.

My work spans embedded systems and controls (VCU programming and CAN bus telemetry for Princeton Racing Electric), autonomous vehicle perception (SLAM, depth estimation, and object detection for Princeton Autonomous Vehicle Engineering), and applied machine learning research (group-equivariant deep learning at CAB Lab). I've also shipped software at General Dynamics Mission Systems and built sensor data pipelines at Robert Bosch.

B.S.E. Electrical & Computer Engineering
Princeton UniversityExpected 2028
Minors in Statistics and Machine Learning, and Robotics
Princeton, NJ

What I've built

A selection of systems, experiments, and research prototypes. Open any card to see a demo or explore the source.

ProteinArchitect

2025

Agentic research MCP servers for protein science, with a custom PyTorch tokenizer mapping natural language into amino-acid embeddings. PrincetonHacks Y-Combinator winner.

PyTorchMCPDaedalus Labs API

DentalConnect

2025

A cloud-based diagnostic web app for oral disease classification, combining a Roboflow CV model with a custom model trained on 10,000+ labeled dental images. SteelHacks XII 3x winner.

AWS SageMakerRoboflowFastAPIReactMongoDB

Where I've worked

Software Engineer

General Dynamics Mission Systems·
InternshipJun 2026 - Aug 2026

Migrated legacy C systems to a C# interoperability layer and built out testing infrastructure for low-level embedded software.

  • Built a C# interoperability layer to migrate legacy C systems, applying object-oriented design to improve encapsulation, eliminate multi-threaded deadlock through synchronized resource access, and optimize concurrency for improved throughput.
  • Built a unit testing framework for low-level C systems using NI DAQmx libraries and created a GitLab CI/CD pipeline around it, increasing validation coverage and catching regressions before code review.
C#CNI DAQmxGitLab CI/CD

Machine Learning Researcher

CAB Lab·Princeton, NJ
ResearchJan 2026 - Present

Researching group-equivariant deep learning for symbolic music understanding, from MIDI preprocessing through model training and evaluation.

  • Built a group-equivariant deep learning model for symbolic music understanding, using group-equivariant convolutions to make harmonic transformations predictable inside the model rather than learned implicitly from data.
  • Built a full-stack inference pipeline with MIDI pre-processing, PLR-group lifting, custom PyTorch layers, model training, OOD evaluation, and feature-map visualization for composer classification.
  • Reduced shifted-input classification error by 20% versus CNN baselines and tested diffusion-based music generation and foundation-model architectures for group equivariance to improve long-range coherence and controllability.
PyTorchGroup-Equivariant CNNsMIDIDiffusion Models

Electrical Engineering Intern

Robert Bosch·Curitiba, Brazil
InternshipJun 2025 - Aug 2025

Built multi-sensor data pipelines for fuel-injector durability testing and an analytics automation tool that cut documentation time by 70%+.

  • Built multi-sensor data pipelines in Python, pandas, and MATLAB for fuel-injector durability testing, converting thermal, pressure, and velocity signals into lifetime metrics and Wohler / S-N curve-based fatigue analysis.
  • Developed a PyQt6 analytics and report automation tool integrating data visualization, AI-assisted summaries, and document generation, reducing documentation time by 70%+.
PythonpandasMATLABPyQt6

Software Team Lead

Princeton Autonomous Vehicle Engineering·Princeton, NJ
ResearchJan 2025 - Present

Leading development of an end-to-end autonomous vehicle perception pipeline, from raw sensor ingestion through 3D scene reconstruction for planning.

  • Building an end-to-end autonomous vehicle perception pipeline that ingests raw sensor/camera data, prototyping Gaussian SLAM and augmented 3D reconstruction, producing real-time obstacle detections, depth maps, and spatial point-map representations for planning.
  • Combined YOLOv8 object detection, TensorFlow-based classification, and Depth Anything 3 depth estimation with RANSAC-based geometric fitting to transform visual distance input into classified scene representations.
YOLOv8TensorFlowGaussian SLAMRANSAC

Embedded Systems Engineer

Princeton Racing Electric·Princeton, NJ
ResearchSep 2024 - Present

Programming the vehicle control unit for the team's electric race car and building a real-time analytics pipeline for the driver dashboard.

  • Responsible for the backend programming of Jetson Orin VCU for data collection and output; used Python and SciPy to implement systems-based vehicle optimization with PWM-controlled cooling and throttle control.
  • Developed a concurrent CAN listener and REST API that streams low-latency vehicle analytics to a driver dashboard.
PythonSciPyJetson OrinCAN Bus