Rohan Gangakhedkar

See the work

Experience

Senior R&D EngineerNov 2025 — Present

EyebotBoston, USA

Designed and shipped the eyescanner's pupillary distance (PD) measurement system from scratch on stereo-depth cameras, building an ML calibration model that maps camera geometry to a pixel-to-millimetre scale; achieved 0.57 mm mean error, a ≈3× accuracy improvement, deployed fleet-wide.
Developed the facetracking algorithm — real-time face detection and 3D localization on Luxonis OAK-D stereo cameras — driving closed-loop linear and angular motor control to align the optics to a user's eyes, with an open-loop feed-forward stage when camera feedback is unavailable.
Built the world's fastest binocular lensometer from scratch, including the core lens border-detection and progressive-lens detection algorithms, extracting SPH, CYL, AXIS and PD from a user's glasses.
Developed the infrared photoretinoscopy eyescanner pipeline (pupil detection, slope extraction, real-time prescription) and a diopter calibration curve that cut prescription error ≈30% below the prior production system.
Created an internal factory test suite through which all kiosk calibration and bring-up is performed — now standard across manufacturing; owned the full release lifecycle.

R&D EngineerMar 2024 — Nov 2025

EyebotBoston, USA

Optimized lens-scanning software in Python for a significant increase in processing speed, and implemented calibration routines for eye- and lens-scanning hardware.
Designed and fabricated mechanical calibration “eyes” via rapid prototyping, and built a reusable repeatability-testing framework used to validate the eye- and lens-scanning fleet.

R&D Engineering InternSep 2023 — Mar 2024

EyebotBoston, USA

Performed data analysis on eye data from clinic operations to identify and correct errors, enhancing the accuracy and reliability of Eyebot's equipment.

Robotics Research Assistant2022 — 2023

New York University · AI4CE LabNew York, USA

Designed and integrated a PCB with an NVIDIA Jetson to unify the systems of a mobile 3D-printing robot.
Managed the design and creation of a multi-DOF robotic arm, enabling optimized control for 3D printing.
Built a simulation environment in Gazebo, Python and ROS to validate the systems of the robot.

Prescription Calibration Pipeline

Eyebot

Built a mechanical-eye calibration pipeline that generates a calibration table mapping the device's raw measurements to true prescription, using monotonic interpolation.
Diagnosed a systematic bias and developed a noise-suppression smoothing method; validated on a 175-eye clinical study (0.366 D mean error, ≈30% better than prior production) and shipped as the live calibration.

LoFTR Feature Matcher Reimplementation

New York University

Re-implemented a state-of-the-art machine learning based image feature matching algorithm (LoFTR) from PyTorch into TensorFlow.
Developed a custom training loop and custom loss function to implement gradient descent on the model.
Implemented a novel transformer architecture to improve the feature extraction of image pairs.

Master of Science, Mechatronics and Robotics2023

New York University — GPA 3.97New York, USA

Bachelor of Mechatronics Engineering (Honours) w/ First Class Honours2020

The University of AucklandAuckland, New Zealand

Programming: Python
Computer vision & ML: OpenCV, PyTorch, TensorFlow, object detection & segmentation, stereo depth, camera calibration
Hardware & tools: NVIDIA Jetson, Raspberry Pi, Luxonis OAK-D, RS485/I²C/UART, Google Cloud, Git