Kunal Kumar Sahoo
I am an undergraduate student exploring the fascinating intersection of Artificial Intelligence and Robotics. My interests lie in developing intelligent systems that can perceive, learn, and interact with the physical world. I am particularly passionate about advancing the fields of Representation Learning, Reinforcement Learning, Computer Vision, and Human-Centered AI.
Beyond my academic pursuits, I actively share my learning journey and technical insights through my blogs, aiming to make complex concepts more accessible to fellow enthusiasts. I am driven by the vision of creating AI-powered systems that can meaningfully contribute to solving real-world challenges.
Contact Me:
|
Experience
- I am currently leveraging Generative AI technologies to develop AI-enabled solutions at S.S.B.I. Digital.
- Formerly, I was also working in the R&D team as intern at Jio Platforms Limited, and PhysioAI.Care where I worked in the areas of 3D Shape Reconstruction and Real-time Pose Analytics respectively.
-
I am associated with several research groups which are as follows:
Lab/Research Group |
Faculty Mentor |
Research Area |
Tenure |
DREAM Lab (CDS), IISc Bangalore |
Dr. Yogesh Simmhan |
Systems for Edge AI |
January 2025 - Present |
Robotics Lab, IIT Gandhinagar |
Dr. Harish P. M. |
Reinforcement Learning for Robotics |
August 2024 - December 2024 |
VISTA Lab (RBCCPS), IISc Bangalore |
Dr. Punit Rathore |
Reinforcement Learning for Traffic Optimization |
May 2024 - August 2024 |
Next-Gen Computing Lab, PDEU Gandhinagar |
Dr. Shakti Mishra Dr. Debabrata Swain Dr. Chintan Bhatt |
Deep Learning for Computer Vision |
August 2023 - December 2024 |
Speech Lab (ATDC), IIT Kharagpur |
Dr. S Das Mandal |
Deep Learning for Speech Processing |
May 2023 - June 2023 |
Education
Degree |
School/University |
Percentage/CGPA |
Duration |
B.Tech. in Computer Engineering |
Pandit Deendayal Energy University, Gandhinagar |
9.68 / 10 |
2021 - Present |
Higher Secondary Education |
Kendriya Vidyalaya No. 1, Gandhinagar |
94.8% |
2021 |
Secondary Education |
Kendriya Vidyalaya No. 1, Gandhinagar |
94.6% |
2019 |
|
Skills
Programming Languages |
Python3, C/C++, Julia, Core Java, Bash |
Database & Platforms |
MySQL, SQLite, MongoDB, Google Cloud Platform, Google Firebase, HuggingFace |
Libraries/Frameworks |
PyTorch, TensorFlow (Keras), OpenCV, PIL, FastAI, PySpark, Flask, Streamlit, LangChain, PyBullet, Mujoco |
Developer Tools |
Git/GitHub, GNU/Linux, Docker, Jupyter, VMWare |
Soft Skills |
Executive Presentation, Project Management, Public Speaking, Team Collaboration, Time Management |
|
Publications & Patents
- P Kadambi, A R Saikiya, P Gupta, K Sahoo, A Joshi, Y Simmhan. Pagoda: Roofline Characterization of Energy and Time for DNN Inference and Training on Edge Accelerators. Submitted to ACM International Symposium on High-Performance Parallel and Distributed Computing, 2025.
- K Sahoo, D Pramanik, P Bharati, S Chandra, S Mandal, T Bhowmik. Explainable AI Driven Deep Learning for accurate L1 Identification from L2 Speech. Presented at 28th International Symposium on Frontiers of Research in Speech and Music, 2024.
- T Patel, A Nour, K Sahoo, A Bostani, D Swain, B Acharya, D Singh. PoseCor: Integrating Sustainable Development and Health Care 4.0 with Deep Learning for Real-Time Pose Detection and Correction. Submitted to IEEE Access, under review.
- S Mishra, C Bhatt, A Patel, K Pipariya, K Sahoo, S Mehta, S Vyas. Modular Device for Multi-Terrain Autonomous Driving for Unmanned Vehicles. Provisional Product Patent published, Applicaion No.: 202321038021
- T Patel, K Sahoo, K Pipariya, D Swain, D Parikh. IoT Based Intelligent Posture Monitor. Provisional Design Patent published, Application No.: 202321037990
|
Projects
Reinforcement Learning for Manipulation Skill Learning
Developed Deep Reinforcement Learning Policies for downstream object manipulation tasks like in-hand rotation, object grasping, and manipulators for precise torque controlled tasks.
Technologies: Python, PyTorch, Stable Baselines 3, Mujoco, ROS
Courtesy: IITGn Robotics, Pandit Deendayal Energy University
Reinforcement Learning for Traffic Congestion Optimization
Modeled city road networks as graph data structure, and deployed multiple reinforcement learning agents to control traffic light signals based on traffic density to minimize queue length and average waiting time of vehicles.
Technologies: Python, PyTorch, Stable Baselines 3, Gymnasium, SUMO, Open Street Maps
Courtesy: Indian Institute of Science Bangalore, Bharat Electronics Limited Bangalore
Swarm Drones Formation
Developed swarming-inspired control algorithms for making geometrical formations by 3-5 drones. Developed Gazebo simulations to demonstrate the working of our algorithm.
Technologies: ROS-Noetic, Gazebo, C++
Courtesy: RoboFest 3.0 by Gujarat Council of Science and Technology
Automated Obstacle Detection and Localization System
Developed an end-to-end pipeline to detect obstacles from vehicle dashboard camera feed using fine-tuned object detection model and then leverage depth-estimation models to generative a bird-eye view of the the scene with respect to our vehicle for enhanced motion planning purposes.
Technologies: Python, OpenCV, PyTorch, YOLO, MiDAS, ROS-Noetic
Courtesy: Pandit Deendayal Energy University Gandhinagar, Military Institute of Armament Technology Poland
3D Bounding Boxes on Dashboard Camera Feed
Developed a custom neural network that utilizes a U-Net backbone to estimate the depth of vehicles present in the dashboard camera feed of the KITTI dataset. Fused the height and width information of vehicles achieved from YOLO algorithm and the depth estimation from our custom neural network to draw bounding boxes. Secured runner-up position in EnCode hackathon at IITG sponsored by Bosch.
Technologies: Python, OpenCV, TensorFlow, YOLO
Steering angle determination using Computer Vision
Developed a simple machine learning application to segment road using K-Means Clustering and determine steering angle for curved roads using pixel density of segmented results.
Technologies: Python, OpenCV, Scikit-Learn
IoT Based Intelligent Posture Monitor
Developed an IoT-based solution for real-time posture monitoring using computer vision and machine learning. The system uses image filtering and skeletonization techniques to detect key body points, followed by mathematical modeling to calculate sitting angles. Implemented multiple ML models (SVM, MLP, KNN) achieving 98.14% accuracy in posture classification. The solution was deployed on edge devices (Raspberry Pi, Jetson Nano) for real-world applications.
Technologies: Python, OpenCV, TensorFlow, MediaPipe, Jetson NANO
Courtesy: Pandit Deendayal Energy University Gandhinagar
Micromouse robot
Developed a compact, wheeled-robot to explore a given maze and learn to solve it. The robot uses 3 ultrasonic sensors to perceive its environment and identify open paths and walls. The maze is learned and solved using the Flood-Fill algorithm which was deployed on an Arduino Nano microcontroller. To achieve precise motion, N20 motors with encoders were used in a differential drive setting along with PID control.
Technologies: Arduino Nano, C++
|
Workshops & Talks
Machin-o-logy @ Parul University
A hands-on workshop on introduction to Deep Learning and it's application in various domains organized by Google Developer Students Club Parul University.
February 2024
Advanced Python Programming Workshop @ PDEU
Hands-on workshop on advanced Python concepts conducted by ACM PDEU Student Chapter. Students got to learn and use Python libraries for web scraping, task automation, regex querying etc.
February 2024
Unsupervised Learning Workshop @ PDEU
An introductory workshop on Unsupervised Learning conducted by Encode: The Computer Science Club of PDEU where students learned the fundamentals of unsupervised learning tasks like Clustering and Dimensionality Reduction. Students also developed a project of image compression using K-Means Clustering algorithm.
January 2024
Arduino 101 @ PDEU
An introductory hands-on workshop conducted by Cretus: The Robotics and Automation Club of PDEU. Students learned the fundamentals of Arduino Programming and at the end of the workshop developed a Line-Following robot from the concepts they learned.
April 2023
|
|