Experience

Nimble AI

Nimble builds autonomous industries powered by industry-generalist superhumanoids.
We developed the world's first general-purpose warehouse superhumanoid capable of performing all core warehouse tasks including storage, retrieval, picking, packing and sorting.

Head of AI

Oct 2019 - current

Nimble AI: Foundation Model for Robotic Item Handling

Powering autonomous picking and high-density packing for millions of unique items

Website

Intelligent Warehouse Orchestration

Collaborative robot fleet coordination with multi-agent RL and resource-aware optimization

Website

AI-Powered Anomaly Detection and Order Verification

Leveraging Vision-Language Foundation Models to Ensure Order Accuracy in High-Speed, Automated Fulfillment Systems.

Website

Samsung R&D Institute

Senior Software Engineer

Jul 2014 - Jul 2017

• Obstruction-free Photography
  Implemented a mobile app for obstruction-free photography—removing reflections/fences from short handheld bursts—and engineered the pipeline for near real-time on-device performance, building on established techniques in computational photography.
• Person-centric Image Gallery
  Built a person-centric photo gallery—similar to the face-grouping feature common on mobile phones—using learned face embeddings and vector search to cluster identities and fetch all photos of a selected person.
• Real-time Face Appearance Modification
  Developed a real-time face filters/beautification stack using deep-learning 3DMM fitting for geometry and lighting estimation, enabling high-quality relighting, and gentle reshaping on mobile.
• Continuous Swipe Input Recognition for Smart Keyboard
  Designed a system to interpret swipe-typing gestures, which works by analyzing both the shape of the swipe and the likelihood of each letter as predicted by a learned language model.

Education

Carnegie Mellon University

Masters in Robotics

Advisor: Prof. Deva Ramanan.
Thesis: Recognizing tiny faces.

Indian Institute of Technology, Bombay

Bachelors of Technology in Electrical Engineering w/ Honors

Publications

Recognizing Tiny Faces

Recognizing faces in low resolution surveillance videos

Website

Low Dimensional Deep Features for facial landmark alignment

Robust facial landmark prediction to enable photorealistic face editing

Website

Hierarchical deep learning architecture for 10k objects classification

Website

Personal

Outside work, I love hiking, running, and traveling—and stress testing my emotional resilience by supporting Manchester United.

Hosted on GitHub Pages — Theme by orderedlist