SOUMYADEEP CHOUDHURY
- |-| Github/Projects
www.linkedin.com/in/soumyadeepchoudhury
sites.google.com/soumyadeep_research_website
EDUCATION_______________________________________________________________________________________________________________
The University of Texas at Dallas, Texas
Aug ‘19 – May ‘21
Master of Science (M.S.) in Computer Engineering (Embedded Systems Concentration)
GPA - 3.52/4
Relevant Coursework: Algorithms, Machine Learning, Distributed Systems, Deep Learning, Computer Vision, Operating System, Database, ASIC
Maulana Abul Kalam Azad University of Technology, India
Bachelor of Technology (B.Tech.) in Electronics and Communication Engineering
July ‘11 – June ‘15
GPA - 7.94/10
TECHNICAL SKILLS__________________________________________________________________________________________________________
Programming Languages: Python, C++, Java, Linux/Unix, ROS, GPU, C, SQL, SPARQL, Verilog, Rest/Client API, WebGL, Embedded-C, CAN, TCP/IP, CSS
Tools & Libraries: Numpy, Pytorch, TensorFlow, Scikit-Learn, OpenCV, Gazebo, Rviz, Moveit, Django, XML, Solid Works, Xilinx, Cadence, Cuda/Cudnn
WORK EXPERIENCE_________________________________________________________________________________________________________
TCS Research & Innovation: Robotics & Deep Learning Researcher: Delhi/Kolkata, India
Oct ‘15 – June ‘19
Amazon Picking Challenge - Picking/Stowing Task: RoboCup 2017, Japan
●
Solved the Space Packaging Problem using Depth Valley Estimation & calibrated the depth camera to optimize the sensory fusion data
●
Refactored the Region-based CNN to track the occluded objects which improved the Object Detection accuracy from 85.4% to 91.2%
●
Honored with prestigious “Special Initiative Innovation Award” for my research contributions, which led the team to 3rd Rank
Managing a Fleet of Autonomous Mobile Robots using Cloud Robotics Platform: IEEE ECMR ‘17; Citations-19: Procter & Gamble
● Invented a Global Planner algorithm to compute an optimal path for each robot based on its current and feedback states in ROS Cloud
● Led a team to optimize the Gmapping noise parameters in SLAM that improved the navigation of Turtlebot in a real 3D environment
● Constructed complex robotic 3D models with detailed physical properties that can be rendered in ROS Gazebo World
Distributed Semantic Knowledge Framework for Collaborative Robotics: IEEE PerCom ’19; 2 Indian Patents (filed)
● Designed an AI model to infer semantic data and execute a complex task by exchanging queries among robots in a poor network
● Outperformed the centralized system in Search and Rescue scenarios by a margin of 45% and honored with the “GEMS Innovator Award”
Vehicles/Trucks Container Number Recognition: Dubai Ports Corporation
● Automated the Truck Management System by developing a RaaS application reducing the congestion up to 15% at Terminal Operations
● Applied Spatial Transformer Network on real-time images captured from cameras to detect 4 container numbers in less than 0.25 ms
Autonomous File Delivering Robot: Marriott Hotel
● Expedited the administrative processes & hotel services by 26% by developing a robot to transfer payloads in a mapped environment
● Prototype new perception capabilities with visual servoing in robots - P3Dx, Fanuc CRX Cobot in simulated and bench environments
INTERNSHIPS & RESEARCH ASSISTANSHIP_______________________________________________________________________________________
Vecna Robotics, Fleet Automation: Graduate Robotics Software Intern: Waltham, MA
June ‘20 – Aug ‘20
● Distributed the tasks among robot groups & PLC systems using a data-protocol handler to minimize the network latency by 13.7%
● Implemented a SCADA system to synchronize the external signals and convert them to mission goals interpreted by robot servers
● Scheduled the robot tasks in the critical section area by estimating the priority of mission goals using a Regression Model with 96% accuracy
The University of Texas at Dallas, TIES Lab: Graduate Research Assistantship: Richardson, TX
Aug ‘19 – May ‘20
● Compressed the training model size to 40% by pruning the least sensitive computational bits in Graph Convolution Network
● Formulated 2 Trojan-Attack Methods to analyze a permanent fault in DNN Accelerator that diminished the prediction accuracy by 60%
Undergraduate Internships: Autonomous Road Surveillance: Delhi, India
Dec ‘13 – July ‘14
● Developed a robust Convolutional Neural Network model to track damages, curbs, and obstacles in roads that reduced accidents up to 30%
ACADEMIC PROJECTS_______________________________________________________________________________________________________
Fire Fighting Robot : IIT-Kharagpur
● Constructed an Aerial Ground Robot with 6 DoF and grasping capability to traverse, avoid obstacles & rescue objects in hazardous terrain
Humanoid Dance Pose Learning - NAO Robot
● Generated a Multimodal Sensing Model to recognize patterns in human dance gestures that enabled a robot to learn dance moves
Mobile-Assisted Speech Controlled Robot (Thesis)
● Proposed an adaptive speech recognition method to detect the human voice commands and control the robot’s movements
