Perception, NavigationPerception, NavigationPerception, Navigation
& AI
Adyansh Gupta
|
Profile
Background & Expertise
Education
RWTH Aachen University
M.Sc. Robotic Systems Engineering
Vishwakarma Institute of Information Technology
B.Tech. Mechanical Engineering - CGPA: 9.4/10 (German Grade: 1.3)
"TransformingtheoreticalAImodelsintorobust,real-worldhardwaredeploymentsforAutonomousMobileRobotsindynamicenvironments.
"
Tech Stack
Professional Journey
Building intelligent systems, from perception pipelines to autonomous mobile robots.
Robotics Software Developer
Robotics Collective
Aachen, Germany
Developing an automated cable insertion pipeline using a robotic arm and RGB cameras for precise, vision-guided manipulation. Integrated teleoperation and speech-recognition on a Unitree R1 humanoid to enable real-time human-robot interaction for live demonstrations.
Mobile Robotics Intern
Addverb Technologies
Noida, India
Engineered a new C++ implementation of the WhyCode/Whycon fiducial marker system utilizing HPC techniques. Worked on camera drivers using shared memory and assisted in integrating Lidar-Inertial Odometry (LIO) for improved robot positioning.
Robotics Intern
Sakar Robotics
Pune, India
Collaborated on development of autonomous industrial AMRs with advanced navigation systems. Implemented computer vision using AI technologies and worked extensively with NVIDIA Jetson hardware and Isaac ROS.
Drone Systems Development Intern
Avignon University
Avignon, France
Developed drone systems for Indoor Navigation, Object Tracking, and ArUco-based Inventory Management using ROS2. Achieved a perfect score of 20/20, distinguishing myself among all students.
Engineering Design Intern
Dassault Systems
Pune, India
Led the conversion of a Pelton Wheel Turbine into a precise 3D digital twin and performed simulations for optimal performance.
Vice President, Mechatronics Engineer
VIIT Robotics (College Club)
Pune, India
Led a team of talented robotics enthusiasts, fostering collaboration and ensuring efficient project execution. Utilized mechatronics engineering skills to design, prototype, and test robotics components.
Engineering the Future
A showcase of autonomous systems, computer vision pipelines, and robotics research pushing the boundaries of machine intelligence.
Go2 Autonomous Inspection (Unitree Go2)
A simulation-first ROS2 stack that lets a Unitree Go2 quadruped autonomously inspect an unknown facility — driven end-to-end by plain-English commands. The robot maps the area on its own, splits it into rooms ("zones"), then walks each zone and reports what it sees using an open-vocabulary camera detector, all triggered through an MCP server bridged to Claude.
AI+Robotics Hackathon 2025 (RWTH Aachen)
Developed under intense hackathon constraints at RWTH Aachen, this project secured the Best Technical Implementation prize by bridging advanced spatial computer vision with real-time robotic manipulation. The core engineering objective was to enable a robotic manipulator to execute reliable pick-and-place tasks within a cluttered, unpredictable workspace without pre-mapping the environment.
OLIVE: Optimization of Lidar, Inertial, Vision & Encoders
A graph-based multi-sensor fusion backend that fuses LiDAR, tightly-coupled IMU, wheel encoders, WhyCode fiducial markers, and monocular visual odometry into one iSAM2 factor graph for drift-free AMR localization. Publishes two synchronized odometry outputs — a globally-accurate map-frame pose and a smooth, jump-free local stream a Nav2 controller can drive on directly — validated to 2.0 cm RMSE over three 56 m loops.
WhyCode ROS2 Package (Addverb Technologies)
Developed whycode_vision, a high-performance ROS2 package for 6-DOF WhyCon/WhyCode fiducial marker localization. Leveraged C++ and AVX-512 SIMD to slash CPU utilization from 180% to 10-15% while enabling multi-marker decoding, hysteresis-based ID stabilization, and hierarchical triangulation for AMRs.
Sepsis Atlas (Clinical AI Evidence Processing)
Developed a local-first RAG intelligence platform to convert unstructured medical research into schema-validated clinical evidence tables. Integrated hierarchical parent-child chunking with ChromaDB, Pydantic schema-driven data extraction via Claude-3.5-Sonnet, and programmatic source-quote verification to ensure hallucination-free output.
R2 - ABU Robocon 2024
Led the software and systems engineering for "R2," a fully autonomous, four-wheel mecanum drive robot for the ABU Robocon 2024 competition. Engineered a comprehensive ROS2 and Micro-ROS workspace featuring YOLO object tracking, LSA08 line following, and TF-Luna LiDAR alignment.
Kiwi Drive Robot
Engineered a custom, small-scale three-wheeled omnidirectional "Kiwi Drive" robot from the ground up. Bridged physical hardware (ESP32, N20 motors) with high-level ROS2/Nav2 architecture via Micro-ROS, featuring EKF sensor fusion and a CNN-based vision pipeline trained on the LVIS dataset.
Vanguard (eYRC 2023-24)
Built for the eYRC 2023-24 "GeoGuide" theme, Vanguard is an autonomous ESP32-based robot guided by an overhead camera system ("Watchtower"). Integrated a custom CNN model to classify simulated war-zone events, enabling the robot to perform real-time A* pathfinding and QGIS geo-tracking to navigate priority targets while avoiding dynamic danger zones.
AgroBot (Omni-Directional Plucking Robot)
Engineered a ROS2-based autonomous agricultural robot featuring an omni-directional X-drive base, dual ESP32 Micro-ROS firmware, and depth-camera vision. Developed end-to-end mission flows—from YOLO-driven cotton boll detection to Cartographer SLAM navigation and automated manipulator kinematics.
Development of Drone Systems
Developed autonomous navigation systems using the Crazyflie drone with MultiRanger and Flow decks, enabling precise indoor navigation via ROS2, Nav2, and SLAM Toolbox. Worked on Tello EDU drone for various computer vision tasks.
Built to Win
Rapid prototyping under pressure — autonomous agents, robotics pipelines, and AI systems engineered (and ranked) against the clock.
Game AI Hackathon
Developed an autonomous pathfinding agent for the Hex game, outmaneuvering competing bots through efficient search and adversarial strategy.
Bonding × Itestra Hackathon
Built autonomous bots for a multiplayer Snake variant, fusing real-time pathfinding with on-the-fly resource management under tight latency constraints.
AI+Robotics Hackathon 2025
Recognized for a flawless sim-to-real robotic manipulation pipeline — bridging spatial perception (NvBlox, FoundationPose) with collision-free MoveIt planning.
Publications, Patents & Certifications
Academic contributions and professional credentials in robotics and AI
Detection and Classification of Bleeding and Non-Bleeding Frames in Wireless Capsule Endoscopy
Novel approach for detecting and classifying bleeding frames in wireless capsule endoscopy images using computer vision.
Get In Touch
Last login: ... on ttys001
Welcome! Initializing contact protocols...