Divam Trivedi | Robotics Engineer

About Me

I'm a robotics and perception engineer pursuing my M.S. in Robotics at Worcester Polytechnic Institute (expected graduation by May 2026), with a strong passion for building intelligent robots that see, think, and act.

With hands-on experience in end-to-end perception and control systems for mobile and manipulation robots, I specialize in bridging the gap between research algorithms and real-world prototypes. My work spans computer vision, sensor fusion, deep learning, and embedded deployment on platforms like Jetson.

I work best on projects that combine algorithms, hardware, and real-world impact. I thrive where hardware meets algorithms, and where perception drives autonomy.

If it involves Robots sensing, deciding, or acting - I want to build it!

Perception

2D/3D vision, object detection, pose estimation, photogrammetry, septh fusion

Machine Learning

Deep RL, Imitation Learning, CNNs, Transformers

State Estimation

Visual-Inertial Odometry, EKF/IEKF, SLAM

Control & Dynamics

Kinematics, trajectory planning, computed-torque control

Systems

ROS2, Gazebo, Jetson, MuJoCo, Blender, Docker

Programming

Python, C/C++, PyTorch, TensorFlow, OpenCV

Experience

Research Assistant - Human-Inspired Robotics (HiRO) Lab

Worcester Polytechnic Institute | January 2025 - Present

Building the perception stack for a nursing-assistance robot, including multi-camera calibration, depth fusion, and object recognition.
Developed a vision-only 6-DoF pose estimation system with under 5% error and integrated visual SLAM for indoor navigation.
Created a gesture-enabled accessibility interface (MediaPipe + ROS2) with over 90% recognition accuracy.
Deploying and optimizing perception models on Jetson Nano, achieving ~2.0 s/frame embedded inference.

Research Assistant - Fire Protection Engineering

Worcester Polytechnic Institute | August 2025 - Present

Conducting combustion diagnostics using cone calorimetry, analyzing heat release rates, ignition behavior, and material combustion properties.
Developing stereo-camera Particle Image Velocimetry (PIV) workflows for flow-field measurement, including calibration and vector field reconstruction.
Designed DAQ interfaces for multi-channel thermocouples and heat flux gauges using LabView, enabling synchronized high-rate thermal data acquisition.

Research Intern

Indian Institute of Science (IISc) | November 2023 - June 2024

Integrated camera-LiDAR fusion and NeRF perception modules with ROS2 control pipelines on NVIDIA Jetson Nano
Executed SfM pipeline achieving <5% geometric deviation against ground truth
Improved navigation accuracy by 20% through multi-sensor calibration for robust SLAM
Optimized 3D model generation algorithms, reducing processing time by ~15%

Machine Learning Intern

Nokia Corporation | October 2023 - February 2024

Engineered digital twin framework for 4G/5G networks, improving PRACH detection accuracy by ~20%
Built real-time KPI monitoring pipelines reducing latency by ~15%
Deployed ML models for dynamic network optimization, improving cell traffic management by ~25%

CTY Mentorship Program - Security Lead

Hewlett Packard Enterprise (HPE) | February 2023 - June 2023

Led detection and mitigation of runtime security vulnerabilities in Kubernetes clusters
Reduced security exposure by ~40% and improved detection time by 25%
Built documentation and taught workshops on container security

Featured Projects

Nursing Robot Perception Stack

Built comprehensive perception system with multi-camera calibration, depth fusion, and object recognition for bi-manual mobile nurse assistant. Achieved <5% pose estimation error for medical objects.

ROS2 PyTorch Jetson Nano Intel RealSense

6-DoF Pose Estimation

Developed marker-less pose estimator for medical objects using vision-only pipeline and multi-view fusion. Eliminated physical markers, reducing setup time by ~30%.

Python TensorFlow Computer Vision

RL Control for UR10 Robot

Implemented DDPG and A3C from scratch for manipulation tasks. Built imitation learning pipeline for peg-in-hole tasks with action chunking transformers.

PyTorch Reinforcement Learning Transformers

Visual-Inertial Odometry

Created >10,000 frame synthetic Vision+IMU dataset. Implemented MSCKF filter and deep fusion network achieving <5% trajectory recovery error on benchmarks.

Blender SLAM Sensor Fusion

Gesture-Controlled Robot

Built simulation and control stack for non-verbal gesture control. Achieved >90% gesture detection accuracy and reduced adaptation time by ~40%.

ROS2 Gazebo MediaPipe OpenCV

Autonomous Vehicle Dashboard

Developed 12 FPS dashboard with lane detection, object tracking, pedestrian pose estimation, and collision prediction for Tesla Model S video streams.

YOLO RAFT OpenPose Blender