Ritam MukherjeeRésumé

I’m Ritam Mukherjee, 2A Tron @uwaterloo

Lately I'm deep in

Fig. 01 · osoto gari entry, reconstructed from videotori red · uke ink · 6.6 s loop · drag to rotate
Incoming

Research Assistant

Demonstration-Based Robot Learning for Industrial Manipulation

January 2027 · Ottawa, Ontario

Now
  • Robotics Engineer @ WATonomous, imitation learning on a simulated humanoid
  • Perception researcher @ WATiMAKE, 3D reconstruction of judo throws
Study

B.A.Sc. Mechatronics Engineering

Expected May 2030

Click a row to read it
  • 2025ParcelCollaborative code editorRepo
  • 2024Recursive Maze SolverPathfinding visualizerRepo
Current thinking
  1. 2026.11

    Everyone's scaling the policy, nobody's checking the clock

    Most of the policy work I read leads with model size and demonstration count. Building the WATonomous ingestion pipeline, I found my episodes were silently corrupting from dropped messages on an under-frequency publisher, and the data looked fine until I checked it. I think demonstration integrity is a bigger bottleneck than parameter count right now, and it stays invisible in benchmarks because the datasets are assumed clean.

Robotics
ROS 2MuJoCoGazeboLeRobotMoveItFoxglovePath Planning (A*)Pure-Pursuit ControlKinematics (FK/IK)State Estimation (EKF/UKF)TeleoperationSolidWorks
Learning
PyTorchImitation LearningACTQLoRA / LoRAQuantizationHuggingFaceRTMPosePose2SimOpenCVNumPyFFmpeg
Languages
C++PythonTypeScriptJavaScriptSQLBash
Systems
FastAPIPostgreSQLRedisTemporalDockerAWSCI/CDpytestLinuxGitNext.js

Open to robotics and research work. Email is fastest.

or mukherjee.ritam@outlook.com