Shreyash Choudhari

Implemented a VLM-based robotic perception pipeline where audio input was used to identify a target object, perform object recognition and 3D object localization, and enable the robot to search for that object across a multi-floor building and navigate close to it autonomously. Developed landmark-based navigation with zero-shot object detection and prompt-based segmentation, enabling quadruped, legged, and humanoid robots such as Go2 and B2-class platforms to approach target objects using a monocular camera with nearly 20 cm accuracy. Built 2-finger pick-and-place pipelines for humanoid robots using Vision-Language-Action (VLA) models for task understanding and manipulation execution.

RL Engineer at Kalyani Group / Bharat Forge (2025-12 – Present)

• Implemented a VLM-based robotic perception pipeline where audio input was used to identify a target object, perform object recognition and 3D object localization, and enable the robot to search for that object across a multi-floor building and navigate close to it autonomously.
• Developed landmark-based navigation with zero-shot object detection and prompt-based segmentation, enabling quadruped, legged, and humanoid robots such as Go2 and B2-class platforms to approach target objects using a monocular camera with nearly 20 cm accuracy.
• Built 2-finger pick-and-place pipelines for humanoid robots using Vision-Language-Action (VLA) models for task understanding and manipulation execution.
• Implemented monocular-camera-based VLA pick-and-place for a UR5 arm using known-object detection and segmentation with a monocular camera for grasping and placement.

AI/Robotics Engineer at Simplia AI (2025-10 – 2025-12)

• Developed a humanoid-based restaurant management system integrating people tracking, customer flow understanding, and order management.
• Built preliminary simulation pipelines for humanoid-assisted restaurant operations covering inventory handling and order workflow management.
• Contributed to service-robot workflow design for hospitality automation, tracking, and operational coordination.

Research Associate – AI/ML & Robotics at IISc Bangalore (2024-07 – 2025-10)

• Developed a mission planner for a plane carrying out imaging using Electro Optical camera, IR camera, and Synthetic Aperture Radar imaging sensors.
• Optimized imaging flight paths for multiple targets based on sensor ground coverage and imaging requirements.
• Project submission completed; currently being patented by the senior team.
• Developed an edge-deployable model on NVIDIA Jetson AGX for few-shot and zero-shot generalization using test-time adaptation with VLMs, improving adaptability in diverse environments.
• Targeted domain generalization and sim-to-real adaptation by fine-tuning VLMs and using TRL-based supervised fine-tuning, improving real-world applicability.

Systems Engineer at EMERGING TECHNOLOGY (2023-02 – 2024-06)

• Developed ADAS, remote control, follow-the-car, and overtake scenarios and implemented them on a Honda car.
• Developed a package from scratch in Python to control industrial servo motors using Modbus RTU.
• Integrated multiple sensors and actuators including encoder, ultrasonic sensors, and acceleration/brake control systems.

Research Intern – under Prof. Arpita Sinha at IIT Bombay (2022-09 – 2023-01)

• Developed a control algorithm for trajectory planning of a level-5 autonomous car in a highway scenario with camera-based navigation using the Oak-D Pro camera, simulated in Isaac Sim.
• Verified the control algorithm in CARLA and implemented it on a 4-wheeler car with various sensors for level-5 autonomy, improving system reliability.
• Received a project referral at Emerging Technology based on this work and worked simultaneously on IIT and IISc projects.

Bachelor of Technology in Mechanical Engineering – BRACT'S Vishwakarma Institute of Information Technology, Pune