Apurv wearing a Meta Quest headset while operating a robotic arm in the lab

Hi! I’m Apurv.

I build intelligent robotic systems.

Robotics engineer working across perception, embedded systems, and human-in-the-loop robotics; turning ideas into tested, working systems.

How I got here.

Working in Prof. Zhu-Tian Chen’s lab, developing human-in-the-loop systems for scalable robot demonstration collection and learning-based manipulation.

  • Building multimodal data-collection pipelines that synchronize RGB-D observations, robot state, and human demonstrations for bimanual manipulation.
  • Preparing datasets and model-training workflows for imitation learning, vision-based policy evaluation, and simulation-to-real validation.
Robot LearningData CollectionModel TrainingHuman-in-the-LoopSim-to-Real

Graduate training at the intersection of perception, autonomy, and intelligent physical systems, with an emphasis on translating algorithms into reliable behavior in simulation and on real robots.

  • Relevant coursework: Computer Vision, Intelligent Robotic Systems, Artificial Intelligence I & II, Machine Learning, and Mixed Reality Systems.
  • Building a systems-level foundation in sensing, learning, planning, and control for reliable deployment on physical robotic platforms.
Computer VisionRobot LearningPlanning & ControlSimulation-to-Real

New Delhi

Led end-to-end product development of a smart medical device, translating user and stakeholder needs into product requirements across mechanical design, electronics, embedded firmware, and prototype validation.

  • Managed the product-development lifecycle across requirements definition, concept iteration, rapid prototyping, subsystem integration, and stakeholder demonstrations.
  • Built and integrated CAD, 3D-printed parts, sensors, electronics, and ESP32 firmware into a working prototype presented to investors and hospitals.
  • Designed 24-hour reliability tests that improved signal stability and reduced measurement inconsistency by 87%.
Product DevelopmentProduct ManagementMedTechEmbedded SystemsPrototype Validation

New Delhi

Developed on-vehicle air-pollution mitigation technologies for new vehicles and retrofit applications in collaboration with IIT Delhi, contributing from system design through prototype validation.

  • Coordinated a team of more than four interns working across mechanical, electronics, sensing, and software subsystems.
  • Contributed to product management and design for manufacturing (DFM), including subsystem integration, testing, and technical documentation for real-world vehicle deployment.
  • Contributed simulation and validation work used in reports that helped secure a $50K+ grant.
Vehicle EmissionsProduct ManagementDFMSystem IntegrationTechnical Leadership

Guwahati, India

Built an interdisciplinary foundation across physics, computation, electronics, modeling, and optimization that now informs my work in robotics and intelligent physical systems.

  • Relevant coursework: Deep Learning; Big Data Analytics: Tools & Techniques; Simulation Techniques in Physical Systems; Optimization Methods in Engineering; Signals and Systems; Digital Electronics and Microprocessors; and Measurement Techniques.
  • Served as a Core Team Member of the Robotics Club at IIT Guwahati, contributing to hands-on robotics projects, peer learning, and technical activities.
  • Combined analytical coursework with laboratory and project experience spanning electronics, sensing, simulation, and computational methods.
Engineering PhysicsDeep LearningSimulationEmbedded SystemsRobotics Club

What I’ve been building lately.

01Robotics research · Teleoperation

Trossen WidowX AI Teleoperation Stack

Two human-in-the-loop interfaces for the WidowX AI, exploring responsive Cartesian control through handheld controllers and vision-based joint mimicry through a physical leader arm.

01Bimanual VR TeleoperationBuilt a Meta Quest 3 teleoperation platform for two WidowX AI arms with ~90 Hz control, estimated sub-35 ms end-to-end latency, calibrated Parallel/Mirrored modes, collision safety, and selectable RGB-D views.

Validated over 8,084 physical control cycles with zero IK or safety failures. Ready for demonstration collection and vision-based policy training.
Watch demo
02Vision-Based Joint MimicryBuilt a vision-based joint-mimicry system that only requires a fully 3D-printed leader arm (with zero electrical or metal parts) and a depth RGB-D camera. Articulated CAD-mesh tracking recovers the leader’s joint configuration directly from the camera feed and maps it to the WidowX AI follower.

Validated the full perception-to-control loop in MuJoCo and on physical hardware, creating an encoder-free interface for demonstration collection and safer joint-level teaching.
Watch demo
Figure 2 flowchart showing the end-to-end Quest and laptop interaction flow

02Mixed reality · Computer vision

Meta Quest 3 Bowling Ball Tracker

A mixed-reality bowling analysis pipeline that sends Quest camera frames to a nearby GPU, tracks the ball through the shot, reconstructs its motion in lane coordinates, and returns a compact result for in-headset replay.

Perception
Classical seed detection + live SAM2 tracking
System
Quest capture → laptop analysis → headset replay
Output
Lane-space trajectory and shot statistics
Semantic energy maps showing depth, saliency, composition, and final retargeted outputsSemantic signals and retargeted outputs
Pipeline from composition, depth, and saliency analysis through energy-map construction and retargetingSemantic-aware retargeting pipeline

03Computer vision · Computational photography

Semantic-Aware Image Retargeting

A class-aware content-retargeting pipeline that adapts its energy function to image composition. DINOv3 predicts composition class, Depth Anything 3 and RGB-VST expose geometry and visual saliency, and a multi-operator stage combines cropping, scaling, and depth-aware seam carving to preserve important content.

Semantics
DINOv3 composition classification
Spatial cues
Depth Anything 3 + RGB-VST saliency
Retargeting
Crop, scale, and depth-aware seam carving

Other technical work

...and when I'm not working on nerdy stuff

Looking for robotics,perception, embedded,or product-developmentroles? Let's talk.

Made by Apurv with Codex · Minneapolis, MN