Uday Sankar

I'm a Robotics Engineer

About

Computer Vision & Robotics Engineer.

Passionate about computer vision, robotics, and deep learning, and I am actively seeking opportunities to apply these skills across diverse domains.

My name is Uday Sankar, and I earned my Master's degree in Robotics Engineering from Worcester Polytechnic Institute in May 2024. I have a strong interest in 3D Computer Vision, particularly in 3D reconstruction and Simultaneous Localization and Mapping (SLAM).

Years of Experience

+ Projects

Positions held

Skills

Programming Languages

  • C++
  • Python
  • MATLAB
  • Bash

Technologies & Tools

  • CMake
  • ROS/ROS2
  • Isaac Sim
  • Gazebo
  • Docker
  • Git
  • Blender

Frameworks

  • PyTorch
  • TensorFlow
  • Keras
  • OpenCV
  • Pandas
  • Point Cloud Library
  • OpenGL
  • g2o
  • GTSAM

Deep Learning Architectures

  • CNN
  • RNN
  • Autoencoder
  • VGG16
  • ResNet
  • DenseNet
  • NeRF
  • Transformers
  • LSTM
  • GAN

Resume

I am actively searching for full-time opportunities in Computer Vision and Robotics. The PDF version of my resume can be found here.

Summary

Uday Sankar

Education

MS in Robotics Engineering

Aug 2022 - May 2024

Worcester Polytechnic Institute, Worcester, MA

Relevant Coursework:

  • Computer Vision
  • Deep Learning
  • Machine Learning
  • Motion Planning
  • Robot Control

BTech in Electrical and Electronics Engineering

Aug 2018 - May 2022

Cochin University of Science and Technology, India

Relevant Coursework:

  • Industrial Automation

Professional Experience

Robotic Perception Intern

Aug 2024 - Present

Onki Robotics, New York, NY

  • Compiling a series of CV modes, like Depth Anything, YOLO, DeepSort, etc. for the Jetson Orin Platform
  • Utilizing TensorRT and other techniques to enhance model performance and increase inference speed
  • Improved frame rate from less than 1FPS to about 6FPS, with ongoing efforts to further enhance model efficiency

Computer Vision Algorithm Intern

Feb 2024 - April 2024

UVeye, Boston, MA

  • End-to-end computer vision project to detect wheel misalignment for an automatic vehicle inspection system
  • Measures the toe angle by detecting vehicle rims by fine-tuning Segment Anything Model
  • Curated datasets using CVAT and ClearML, to fine tune SAM

Graduate Student Researcher

Aug 2023 - Dec 2023

Worcester Polytechnic Institute, Worcester, MA

  • Simulated Franka Emika robot in NVIDIA Isaac Sim
  • A state machine based system using both ROS2 and Isaac Sim API to control the robot
  • LLMs for breaking down complex robotic tasks into a sequence of simpler, more manageable actions

Robotics Intern

Jun 2023 - Aug 2023

Orangewood Labs, San Francisco, CA

  • RoboGPT, an LLM based interface for a 6DoF Robotic Arm
  • Perception stacks for the robot using YOLO, CLIP, and multimodal InstructBLIP model
  • Multi-camera system using Realsense D435 and D405 RGB-D cameras at workspace and end-effector

Portfolio

  • All
  • Computer Vision
  • Deep Learning
  • Robotics

xfeatSLAM

Real-time SLAM with Deep Features (XFeat + ORB-SLAM3)

Simple SLAM

A simple implementation of Monocular SLAM using Python.

Structure from Motion

3D reconstruction using multi-view geometry principles.

Neural Radiance Field (NeRF)

PyTorch implementation of NeRF from scratch.

Perception Stack for Autonomous Vehicles

A perception stack based on monocular data collected from a Tesla Model S

Image Panorama Stitching

A python pipeline to stitch input images into a panorama image

Kinematic Path Planning Under Non-Holonomic Conditions

Vehicle parking into compact space under non-holonomic conditions

Language Translation using Seq2Seq model

English to French translator using a seq2seq model

Image Classification using ResNet, ResNeXt, and DenseNet

Classification of CIFAR 10 Dataset

Image Boundary Detection using PB-Lite Algorithm

Simplified version of Probability of Boundary Algorithm

Contact

Address

36 Lee Street, Worcester, MA 01609