Bengaluru, India
Product Manager · Builder · MIT
I build technical products that turn hard engineering into things people actually use — from cloud data platforms at Google to AI-powered hardware I shipped as a founder.
I'm a product manager who likes living where the hard technical problems are. Today I work on Google Cloud, building the data migration and data protection capabilities behind Google Cloud NetApp Volumes. Before that I owned the next-generation roadmap for Amazon FSx for Lustre at AWS — a high-performance file system used for ML training, EDA, and visual effects.
The thread running through everything I do is building from zero. I co-founded Guard-Ex, where we created an AI device to detect drug impairment from the eyes, and Smart Being, an automated vertical-farming system that grows food with 70-90% less water. Along the way I did a deep dive into computer vision and machine learning, shipped hardware prototypes, and earned my master's at MIT — where my thesis used deep learning to make sense of 3D damage inside aerospace composites.
I care about the same things whether I'm writing a PRD or soldering a board: real user problems, honest engineering, and shipping something that works.
School of Engineering & MIT Sloan. Thesis: "Multiclass 3D Segmentation of Progressive Damage in Advanced Composites using Deep Learning" (NECSTLAB). Also completed the MIT Deep Technology Bootcamp (2019).
Built an automated computer-vision traffic-surveillance system as a capstone effort.
An AI device that detects drug impairment from the eyes. I built the computer-vision pipeline for pupil tracking and nystagmus detection, used Fourier analysis to quantify abnormality, and developed the GX-420 prototype piloted with Waterloo police.
Fully automated indoor vertical farming. Sensor-driven nutrient delivery and IoT monitoring grew fruits and vegetables year-round with 70-90% less water. Government-funded and featured on Rajya Sabha TV.
A CNN (U-Net/VGG) pipeline for multiclass 3D segmentation of microscale damage in advanced composites from X-ray CT scans — cutting manual segmentation from ~10 hours to ~2 minutes. Advised by Prof. Brian Wardle and Prof. Nicholas Roy.
An MIT Sandbox–funded hardware startup building a wearable to help form a meditation habit. Led product across the PCB, embedded firmware, enclosure design, and companion iOS app.
A smart two-camera device that recognizes food items inside your fridge and suggests recipes. Built the prototype end to end: capture, on-device CV + Google Vision, IoT upload, and a web dashboard. Pitched to investors.
An edge device that scores the readability of laser-printed text in real time and alarms when print quality drops. Built on BeagleBone Black with a custom Yocto OS and a per-character ML quality model.
Early-stage skin-cancer screening using classic CV feature extraction and a KNN classifier over thousands of lesion images, returning a calibrated confidence score — aimed at places without easy access to specialists.
Turns a photo of a handwritten restaurant menu board into a structured digital menu via image preprocessing, region-of-interest extraction, and OCR (with Google Vision / Azure fallback).
A prototype that computes aerial vehicle density from camera feeds and dynamically reallocates traffic-light timing via Arduino — also detecting zebra-crossing violations, license plates, and helmets.
A virtual assistant that recognizes user emotion from facial expression and responds accordingly — combining a CNN emotion model with a conversational front end.
Always happy to talk product, AI, or building things from zero. Reach me at sudhirjain@alum.mit.edu.