ROBOTICS ENGINEER — PUNE, INDIA
I build autonomous navigation systems — ROS 2 down to register-level firmware — and I have a habit of ending up wherever the project is stuck: the localizer, the mechanics, sometimes the supply chain.
Joined a robotics team as a programmer. The record since: Nav2 autonomy stacks, register-level firmware, national-podium robots, a supply-chain department built from scratch, and a patent at Tata Technologies — whichever the project needed first.
I'm a final-year Electronics & Communication Engineering student at MIT World Peace University, Pune. My home turf is software and robotics programming — I was one of the five engineers who first brought ROS to MIT Tech Team, and I've gone deep across the stack, from register-level firmware on STM32 and ARM Cortex-M4 to CAN networking to full Nav2 autonomy.
The pattern across those four years: I end up wherever the project is stuck. When the team ran short of mechanical members one season, I stepped outside software entirely — took on the mechanical and supply-chain work myself, saw it through, and built a supply-chain system the team still runs on today. When my capstone's visual localization hit the frame-rate ceiling of our cloud GPU, my teammate and I traced the bottleneck to its root and re-architected the localizer around AMCL, solving the real-time problem cleanly instead of abandoning the approach.
That adaptability has compounded: national podiums at DD Robocon, a stint at Tata Technologies where I'm the first named inventor on a patent disclosure that MIT-WPU selected and is funding for formal filing, and demos to clients like Cummins and Brose.
Today I'm a Teaching Assistant at MIT-WPU — invited back by my mentor of four years to train the next batch of the team I grew up in, while pushing two patents and three papers toward publication.
MIT Tech Team & Dept. of EEE, MIT-WPU · w/ Prof. Dr. Saket Yeolekar
Tata Technologies Ltd.
MIT Tech Team — ABU Robocon
Most indoor robots see with stereo or RGB-D cameras plus an embedded GPU — hardware that dominates the bill of materials. My capstone asks: what if a robot could navigate with only a 2D LiDAR and a commodity webcam-grade monocular camera? The answer cuts sensing-and-compute cost by 3–4×.
RGB streams are shipped to Depth Anything 3 running on a cloud NVIDIA L4, the predicted depth is re-injected into ROS bags, and RTAB-Map (ICP) fuses everything into a clean 2D occupancy grid — cloud-grade perception on a budget robot.
Live navigation runs fully onboard through Nav2 — NavFn A* global planning, Regulated Pure Pursuit control, AMCL localization — over 2D LiDAR and rf2o laser odometry. No GPU on the robot at all.
A custom 3-wheel holonomic omni-drive with closed-loop PID on a TI ARM Cortex-M4 completed 8-waypoint autonomous patrols with 0.08–0.15 m goal error across a ~10 m × 9.7 m indoor arena.
Turns Excel specs into production-grade ARXML and CAN DTC driver code via a 4-pass dependency-resolving generator — a 50-signal component drops from ~2 hours of manual work to 1.42 seconds. Patent filing in progress.
Python 3.11 · PyQt6 · openpyxl · cantoolsController-operated robot with automated sub-sequences — capable of dribbling, jumping, and dunking. Full mechanical + control stack now headed for a patent.
ROS 2 · Embedded C · Motor control · CANIn-house odometry implemented for competition robots in 2024 — now being written up for journal submission as part of the team's research push.
STM32 · Encoders · Kalman filtering · ROS 2
Developed at Tata Technologies — first named inventor. MIT-WPU selected the disclosure and is funding the formal filing: institutional backing, not a self-filed application.
★ MIT-WPU SPONSORED FILINGPatent covering the full mechanical and software design of the ABU Robocon 2025 robot.
★ DEPT. OF EEE SPONSOREDConference paper on the ABU Robocon 2024 system, authored as part of the founding five-member ROS group.
Journal paper targeting Q3-rated submission through 2026.
Journal paper on the capstone navigation stack; targeting Q3-rated submission through 2026.
Before I wrote odometry code, I was the odometry. A decade of competitive skating taught me velocity, balance, and path-planning the hard way — on eight wheels.
Part of the Guinness World Record for the largest skating dance — Bengaluru, India, 2015.
Over 60 medals across school, district, and state level — with international competition invitations.
Represented at state level in roll ball — speed skating's full-contact team cousin.
Completed the entire abacus mental-arithmetic program, every level through Grand Master — high-speed computation before I ever touched a microcontroller.
Awarded a scholarship by MIT World Peace University for MHCET performance — 92nd percentile.
[07 / ESTABLISH LINK]
Open to robotics, embedded, and autonomous-systems roles and research collaborations. The fastest route is email — I usually respond within a day.