Senior Robotics & Computer Vision Engineer specialized in humanoid platforms, sim-to-real locomotion policies, and real-time edge perception.
Over 4+ years architecting and shipping production-grade robotics — currently developing bipedal locomotion policies, safety layers, and sim-to-real transfer for the Unitree G1 humanoid at FEV India. Proven track record deploying sub-50ms TensorRT perception pipelines on factory floors, commissioning 7+ robotic platforms at Gujarat Science City, and engineering aerospace vision systems with DRDO/ADA.
About
I'm a Senior Engineer L1 currently at FEV India, actively developing humanoid robotic systems on platforms like the Unitree G1 with Inspire dexterous hands. With 4+ years of hands-on experience, I design physics-accurate sim-to-real transfer workflows, bipedal locomotion policies, and robust hardware abstraction/safety layers that bring complex robots reliably into reality.
My journey started at RGIT's Robotics Club in Mumbai, where I led simulation and trained 50+ students. From there, I contributed to India's first Robotics Gallery at Gujarat Science City — working hands-on with Pepper humanoids, UR10 arms, Da Vinci surgical robots, SCARA systems, and autonomous mobile platforms.
At Wastefull Insights, I scaled from robotics engineer to senior engineer, architecting autonomous waste sorting systems that process 10,000+ objects daily. I refactored Python to C++ for 80% compute reduction, built multi-model YOLO pipelines with DeepStream, and developed analytics frameworks with 91% throughput prediction accuracy.
The theme has been consistent: take messy real-world problems and convert them into reliable, measurable systems.
B.E. Electronics & Telecommunication
RGIT Mumbai (2017-2021) · CGPA: 7.04
Senior Engineer L1 at FEV India Pvt Ltd
Unitree G1 Humanoid & Inspire Hand
Engineering Tenets
Moving robotics from fragile Jupyter notebook prototypes into 24/7 industrial factory deployment requires solving the messy realities of hardware, compute, and physical environments.
Research prototypes in Python break down under sub-30ms industrial control cycles. I refactor critical perception bottlenecks to multithreaded C++ and serialize deep learning models into TensorRT FP16/INT8 engines on NVIDIA Jetson Xavier, slashing compute overhead by 80% while concurrently processing 4 RTSP feeds.
True delivery means engineering for dust, camera vibration, and ambient factory lighting swings. I design industrial GigE line-scan vision setups, integrate synchronized strobe lighting, program Omron PLC safety interlocks, and architect watchdog auto-recovery loops ensuring 99.2%+ operational consistency.
In production pick-and-place, milliseconds directly dictate facility throughput. By mathematically modeling and refining 3-axis gantry acceleration and S-curve motion profiles, I eliminated inertial losses by 15% and boosted picking cycle speed by 40%, coordinating multi-robot cells sorting 10,000+ items daily.
Robotics engineering must produce verifiable economic impact. I built end-to-end telemetry pipelines streaming real-time vision analytics through Kafka and TimescaleDB, training regression models that predict facility throughput with 91% accuracy and slashing manual quality inspection labor by 60%.
Pushing the boundaries of what robots can perceive, plan, and execute.
Exploring simulation environments like Isaac Sim Arena, RoboStudio, GenieSim, and Lightwheel AI (RoboFinals) to generate physics-grounded synthetic data for humanoid policy training.
Developing first-person multimodal demonstration pipelines (egocentric vision, tactile, proprioception) on platforms like RoboLab for dexterous manipulation and imitation learning.
Engineering closed-loop verification benchmarks and diagnostic safety boundaries to evaluate policy stability before running on physical humanoid hardware.
Scaling whole-body coordination on the Unitree G1 bipedal platform with Inspire multi-finger dexterous hands, coupling locomotion policies with compliant manipulation.
Skills
Comprehensive toolkit built through 4+ years of production robotics and computer vision deployment.
Platforms: Unitree G1, Inspire Hand, Pepper, UR10, NAO, Da Vinci, SCARA, AMRs, Quadrupeds
End-to-end robotics pipeline from simulation through perception, control, and production deployment.
Isaac Sim / Gazebo / RViz
Digital twins, physics testing, synthetic data generation
ROS/ROS2, MoveIt, Nav2
IK, trajectories, collision avoidance, SLAM
YOLO, DeepStream, TensorRT
Real-time detection, tracking, classification
PLC, Motor Controllers, Servos
Safety I/O, real-time control loops
Cameras, LiDAR, IMU
Sensor fusion, calibration, data pipelines
Jetson, Docker, Edge
Production builds, monitoring, analytics
Experience
From college robotics clubs to production factory floors — each role shaped how I build reliable autonomous systems.
FEV India Pvt Ltd · Pune
Feb 2026 – Present
Leading humanoid robotics development focused on the Unitree G1 platform equipped with Inspire dexterous hands — architecting physics-accurate sim-to-real pipelines, locomotion policies, and mission-critical safety architectures.
Wastefull Insights · Vadodara
Sep 2023 – Nov 2025
Led software architecture, perception pipeline development, and technical team coordination for autonomous robotic waste management systems.
Wastefull Insights · Vadodara
Aug 2022 – Sep 2023
Contributed to autonomous robotic systems for waste management, focusing on perception optimization and real-time processing.
Engineering Services International · Ahmedabad
Oct 2021 – Jul 2022
Developed exhibits for India's First Robotics Gallery at Gujarat Science City. Hands-on experience with 7+ diverse robotic platforms.
IN2PETA Services · Hyderabad
Jun 2021 – Oct 2021
Developed autonomous pick-and-place mobile robot for cluttered environments, integrating perception, navigation, and embedded control on NVIDIA Jetson.
RGIT's Robotics Club · Mumbai
Jul 2019 – Jun 2021
Led simulation division and mechanical/software integration for national robotics competitions. Mentored 50+ students in ROS, Gazebo, and mobile robotics.
Production systems, national science exhibits, and defence R&D — complete with verified problem formulations, architectures, and quantified outcomes.
Production DeployedReal-time counting, tracking, and SKU classification across 4 concurrent RTSP camera feeds with 99.2% accuracy.
Production DeployedEnd-to-end vision-guided robotic picking cell with multi-robot coordination sorting 10,000+ objects daily.
Production GalleryFully automated robotic café with UR10 manipulator on 7th-axis linear rail executing precision pick-and-place.
Production GalleryHigh-speed competitive robot playing real-time air hockey against human opponents with Kalman trajectory prediction.
Defence R&DVision-based automated landing pose estimation system for aircraft where GPS degrades near ground level.
Defence R&DPhotogrammetry algorithm recovering 6-DOF trajectory of released aircraft stores from high-speed video footage.
CompletedUnsupervised optical QC station flagging microscopic defects on manufactured grinding wheels using Anomalib & DeepStream.
Production GalleryHuman-robot interaction framework with face recognition, visitor engagement, and choreographed behaviors.
Production GalleryAutonomous 3v3 robot soccer coordination with centralized overhead computer vision and dynamic tactical role switching.
Production GalleryRetrofitted surgical robotic arms with contactless Leap Motion gesture control for visitor demonstrations.
Production DeployedAutomated defect inspection for precision automotive components measuring sub-100 micron geometric tolerances.
Award WinningDual UR5 robot arms coordinated via MQTT for autonomous warehouse sorting, packaging, and order dispatching.
Services
Professional consulting and development services for robotics, computer vision, and industrial automation.
End-to-end design and development of autonomous robotic systems for industrial applications — motion planning, control systems, hardware integration, and production deployment.
Custom vision systems, object detection, segmentation, and AI/ML deployment on edge devices. From training pipelines to TensorRT-optimized production inference.
Virtual representations of physical systems for simulation, testing, and optimization. Sim-to-real transfer techniques for seamless virtual-to-physical deployment.
Hands-on robotics and computer vision workshops for students, professionals, and organizations. Topics include ROS, computer vision, and embedded systems.
One-on-one mentorship for robotics projects, code reviews, and guidance on best practices in autonomous systems development.
Technical presentations at robotics conferences sharing insights on production-grade vision systems and AI-accelerated development.
Collaboration on open-source robotics projects, code contributions, and community-driven initiatives to advance robotics technology.
Field notes, profiling breakdowns, and architectural post-mortems from production robotics deployments, bipedal humanoid locomotion, and edge AI vision.
How we train bipedal locomotion policies with randomized friction, motor delays, and ground inclination in simulation to achieve robust zero-shot push recovery on the physical Unitree G1 platform.
Profiling the 195ms latency crisis in a multi-camera industrial perception stack, uncovering memory copy bottlenecks, and restructuring into a 42ms zero-copy C++ pipeline running on edge Jetson hardware.
Designing a 240 FPS puck-tracking pipeline that executes trajectory forecasting, rebound angle calculation, and striker intercept kinematics within 7ms to beat human opponents.
Practical engineering lessons from 24/7 conveyor sorting facilities: positive-pressure cooling enclosures, predictive GPU throttling before thermal trips, and synchronized strobe illumination.
Contact
I'm always excited to connect with fellow engineers, researchers, and innovators. Whether you have questions about robotics, want to discuss potential collaborations, or just want to say hello — I'd love to hear from you!