Projects

Selected Work

01

Drivetrain

FRC 2056 · Mechanical
2056 drivetrain render — swerve base with electronics

A 32" × 28" swerve base on MK4i modules — Kraken X60 drive geared to 15 ft/s, Kraken X44 steering.

Requirements

  • Agile, responsive, and highly controllable.
  • Extremely robust, rigid, and reliable.
  • Low centre of mass (CoM).
  • Must provide rigid structure and mounting of subsystems.

Details

  • Constructed of 3" x 1" x 1/8" aluminum tubing for stiffness and a lower chassis. A 1/16" solid aluminum baseplate acts as a shear plate, offers a mounting surface for electronics, and lowers the overall CoM to improve stability.
  • Maximum chassis size of 32" long by 28" wide to ensure stability when scoring.
  • The corners of the baseplate are bolted, adding extra support to the rivets.
  • The baseplate is fastened directly to the elevator gearbox, and an aluminum angle riveted along the baseplate near the battery prevents sagging.
  • Utilizes Swerve Drive Specialties MK4i modules, selected for their proven reliability, low CoM, and team familiarity.
  • Powered by Kraken X60 motors for a top speed of 15 ft/s. The L2 gear ratio balances speed and acceleration for the game's short cycle distances.
  • Each swerve module steers with a Kraken X44 on a custom shifted adapter plate and a 12-tooth pinion, saving nearly 2 lbs.
02

Autonomous RC Aircraft

TMU Aero Design · Aerospace
Finished autonomous RC aircraft, covered airframe Autonomous RC aircraft structure, uncovered truss airframe

A 5.5 ft wingspan autonomous airframe — FEA-validated and flight-tested.

Highlights

  • 5.5 ft wingspan airframe designed in SolidWorks
    • Built from plywood, aluminum spars, and carbon fibre composites for a high strength-to-weight structure
  • Static structural FEA on the landing gear under landing-impact loading
    • Redesign validated on real landings
03

VitalGroove

MakeUofT · Robotics

An end-to-end sense–think–act robot built on an NVIDIA Jetson Nano in 24 hours.

Highlights

  • Full perception-to-actuation loop on a Jetson Nano
    • Computer vision detects human interaction, a VLM pipeline decides, and motors actuate
  • Brought up the entire hardware stack under a 24-hour deadline
    • Debugged wiring, power, and inference latency to keep real-time control stable under live camera feeds