VPVidhatu Patel

Electrical & Computer Engineering and Physics, University of Washington

Vidhatu
Patel

I write software that helps robots work out where things are: computer vision, sensor fusion and state estimation. Right now I’m on the software team for UW’s Mars rover with Husky Robotics.

See projectsOpen resume (PDF)
400022003D estimateCam 1Cam 2Cam 3
View A: side elevation. Dimensions in mm.BigTrack’s dorm-room test, drawn as a side view. Each camera’s detection becomes a sight line, and the 3D estimate is where the lines agree. Webcams aren’t clock-synchronized, so a fast ball pulls the lines apart; that timing error, not detection, limited the project’s accuracy.

Projects

Sheet 2 of 5Detail views

  1. 3D estimateCam 1Cam 2Cam 3
    Detail ANot to scale

    BigTrack

    Multi-camera 3D object tracking, 2026

    • Built a 3D tracker where any browser with a webcam joins as a camera, running YOLO detection on the GPU in real time.
    • Fused detections on an async Rust server with a C++ factor-graph optimizer that estimates camera poses and 3D paths.
    • Found that cross-device timing and noise modeling, not detection, limited accuracy, and designed fixes for both.

    Built with Rust, C++, GTSAM, ONNX Runtime, WebGPU

    Read the case studySource on GitHub

  2. Swerve module (4×)TurretTag
    Detail BNot to scale

    FRC Team 900 robot software

    Competition robot with swerve drive and turret, 2026

    • Wrote software for a swerve-drive robot with a turret shooter, structured so every match can be replayed from logs.
    • Fused AprilTag vision with odometry for localization; built turret auto-aim that compensates for latency and motion.
    • Set up a physics simulation of the robot and field so drive, vision, and autonomous code were tested before reaching hardware.

    Built with Java, WPILib, AdvantageKit, PhotonVision

    Team code on GitHub

  3. Voice sampleOn-device CNN
    Detail CNot to scale

    NeuroSense

    On-device Parkinson’s screening, 2025

    • Built an Android app that screens for Parkinson’s disease from a short voice recording, running a CNN entirely on the phone.
    • Used transfer learning and augmentation to train on only 81 recordings, evaluating recall rather than accuracy alone.
    • Rebuilt the audio pipeline in Kotlin when the iOS toolchain failed, and verified it against the Python training pipeline.

    Winner, Congressional App Challenge (NC-5), January 2026

    Built with PyTorch, Kotlin, Flutter

    Source on GitHub

Experience

Sheet 3 of 5Revision history

Experience, most recent first
RevDescriptionDate
C

Software team member

Husky Robotics, University of Washington, Seattle, WA

  • Contributing to software for UW’s Mars rover, which competes in the University Rover Challenge.
Oct 2026 – present
B

STEPS4GROWTH clean energy intern

CLEAN Labs, North Carolina A&T State University, Greensboro, NC

  • Built and deployed autonomous-vehicle software in ROS 2 Foxy on Linux using 2D LiDAR and stereo cameras.
  • Trained and optimized real-time detection and tracking models on an NVIDIA Jetson Xavier NX.
  • Developed LiDAR–camera sensor fusion pipelines to improve localization and obstacle detection.
Jun – Aug 2025
A

Captain, lead programmer and lead fabrication

FIRST Robotics Competition, Teams 1533 and 900, North Carolina

  • Led autonomous-navigation software using YOLO detection and fiducial-marker localization; designed trigger-based command flow and tested logic in simulation to cut on-site debugging.
  • Advised the design team on 3-axis CNC manufacturability and built a CAD-to-part pipeline across subteams.
Jun 2022 – Jun 2026

About

Sheet 4 of 5Notes and parts list

I started in FIRST Robotics, captaining Teams 1533 and 900 and writing their autonomous code. Since then I’ve worked on ROS 2 autonomous vehicles, on-device machine learning and multi-camera tracking. The part I keep returning to is estimation: combining noisy sensors into an answer you can trust, and working out why it’s wrong when it isn’t.

Education

  • University of Washington

    Sep 2026 – Jun 2030

    B.S. Electrical and Computer Engineering; Physics (Comprehensive)

    Coursework: CSE 123 Introduction to Computer Programming III, PHYS 122 Electromagnetism

  • North Carolina School of Science and Mathematics

    Aug 2024 – Jun 2026

    High school diploma; weighted GPA 4.50 / 5.00

    Coursework: Linear Algebra (incl. Advanced Topics), Multivariable Calculus, Graph Theory, Mathematical Modeling

General notes: honors and awards

  1. Congressional App Challenge winner

    NeuroSense; Rep. Virginia Foxx, North Carolina’s 5th District. Jan 2026.

  2. Intel AI Global Impact Festival, Global Award

    SightLine: AI navigation glasses for the visually impaired. 2025.

  3. FRC Dean’s List semi-finalist

    FIRST North Carolina District, nominated for student leadership. Mar 2025.

  4. FIRST North Carolina State Champion

    Team 1533; winning alliance at the 40-team State Championship. Apr 2024.

Bill of materials: technical skills
ItemGroupParts
1LanguagesC++, Python, Java, Rust, Kotlin
2ML and visionPyTorch, TorchScript, ONNX Runtime, YOLO, OpenCV, GTSAM, Transfer learning
3Systems and toolsLinux, ROS 2, NVIDIA Jetson Xavier NX, WebGPU, Docker, CMake, Git, WPILib, AdvantageKit, maple-sim
4AreasComputer vision, Robotics, State estimation, Sensor fusion, On-device ML, Embedded systems

Contact

Sheet 5 of 5Title block

Title

Get in touch

I’m looking for research and engineering work in robotics and perception. Email is the best way to reach me.

Drawn by

Vidhatu Patel

Location

Seattle, WA

Email

vidhap@uw.edu

GitHub

github.com/vidhatu

Resume

Download PDF

Dwg no.

VP-2026-01

Sheet

5 of 5

© 2026 Vidhatu Patel