About Me

I'm a computer vision engineer with a passion for building autonomous MAVs and ground vehicles. I'm currently working on object detection and other aspects of vision pipelines for drones and ground robots. My primary experience is in model optimization, with hands on experience in deployment on edge devices.

Technical Skills

ROS1, ROS2, Gazebo, ONNX and ONNXRuntime, C++14, Python3, Tensorflow 2.x, OpenCV 3.x and 4.x, Git and Github

Interpersonal Skills

Adaptability, Problem Solving, Creativity, Humor, Communication

Ongoing Projects

Multithreaded cpp with opencv

Check it out on github!!

  1. Developing a multithreaded cpp application with opencv for real time image processing
  2. Working on adding GPU based acceleration to the application (Using Nvidia 4060 local GPU)
  3. Working with performance analysis tools for edge devices

Neetcode.io solutions

Check it out on github!! Start Date - Present

  1. Working on solving 150 questions from Neetcode.io

Projects

Hands on with ROS 1

• Rigbetel Labs LLP, Pune 3 months

  1. Learnt fundamentals of Robot operating system (Noetic)
  2. CAD designing and post processing in Autodesk Fusion 360
  3. Environment creation and simulation in Gazebo
  4. Implementing SLAM algorithms and utilizing ROS navigation stack
  5. Brief introduction to ROS 2

Prototype of a research quadcopter

•IEEE ProjectJan 2020 - May 2020

  1. Developed a low-cost, open-source hardware stack-based quadcopter using Pixhawk/ APM 2.8 with F450 frame, aimed at robotics researchers and hobbyists.
  2. Led a team focused on component integration, design, fabrication, and market research for the final prototype.
  3. Designed various mounting and accessory parts for F450 frame using Fusion360

Deep learning based autonomous navigation for quadrotors

• Master's Project Aug 2019 - Aug 2020

  1. Developed understanding of deep learning architectures and datasets for building autonomous drones.
  2. Worked with a Jetson Nano-based drone for model deployment, inference workload balance, and optimization.
  3. Proposed a neural network based on pyramidal architrcture 1M parameter for yaw prediction on a small board computer (Jetson Nano)

Geometric vision based navigation algorithms for quadcopters

• Bachelor's ProjectJan 2019 - May 2019

  1. Studied 3D vision-based methods for scene reconstruction, map generation, and navigation

Work

Arka Aerospace Pvt Ltd @ Hyderabad

Computer Vision and Robotics Engineer January 2024 - Present

  1. Constructed inference pipeline for edge devices using Pytorch, TensorRT, and Nvidia's DeepStream library to achieve speeds up to 120 FPS for custom object detection using YOLO models.
  2. Designed object detection and tracking models on edge devices using various YOLO models and custom trackers.
  3. Researched methods in data collection and cleaning for custom classes of objects from web and in-house data sources.
  4. Implemented and optimized computer vision algorithms for autonomous navigation systems.
  5. Collaborated with cross-functional teams to integrate vision solutions into robotics platforms.

Multicoreware Pvt Ltd @ Chennai

Software engineer Feb 2022 - December 2023

  1. Achieved up to 3x performance boost by optimizing models and libraries for accelerated inference on server-grade hardware.
  2. Enabled model addition to the client's model zoo by building testing and benchmarking suites in ONNXRuntime framework.
  3. Developed expertise in CNNs, NLPs, and recommender systems using onnxruntime framework.
  4. Developed automation scripts for performance and accuracy runs for various models.

Bitsilica Pvt Ltd @ Hyderabad

Machine learning engineer Feb 2021 - April 2021

  1. Optimized CNN model's inference for edge devices resulting in the adoption of better-suited multi-device options.
  2. Utilized Django and Flask to improve API functionality for client demos.
  3. Developed a news reader app for extracting headlines from news feeds using OpenCV and PyTesseract.

Avian Aerospace Pvt Ltd @ Bangalore

Robotics and computer vision Intern Dec 2017 - Feb 2018

  1. Designed object detection algorithm for an underwater ROV deployed on an onboard computer for client demo.
  2. Developed algorithms using libraries like NumPy and OpenCV for object detection, line following, and distance estimation.
  3. Desigend the hardware for ground rover, supplemented with code for following a line using camera and light sensors.

Education

Central University of Karnataka

• Integrated Masters in computer Science• CGPA 8.53

Project driven Integrated masters focusing on computer vision and deep learning. Supplemented by my project on building a small CNN for autonomous yaw control of a quadcopter.

Contact Details

Atmadeep Arya
aatmadeepaarya@gmail.com