Professional Experience
Principal ML Scientist
March 2024 - PresentPhronetic AI
Bangalore, India
Working on a low-code multimodal streaming AI platform for developers.
Key Projects:
- AI Agent Builder: Design and implementation of agentic AI workflows for processing text, audio, and video streams
- Real-time Talking Face Generation: Adapted Gaussian Splatting based model for real-time rendering
- ABM (AI Business Manager) Pipelines: AI-based activity monitoring for domain-specific use-cases
- Vision Planner: Video stream processing system design
- Owlet Model: Adapted lightweight vision-language model for video understanding (blog)
Gaussian Splatting Vision-Language Models PyTorch Multimodal AI
Technology Consultant (CV / ML)
Nov 2023 - March 2024Independent
Bangalore, India
Focused on multimodal search in fashion e-commerce, video intelligence platform design, and custom ML research.
Key Projects:
- Multimodal search systems for fashion e-commerce
- Video intelligence platform architecture
- Custom ML research including latent diffusion models
Computer Vision Machine Learning Multimodal Search Latent Diffusion
Team Lead - Vision Team
Dec 2013 - Oct 2023Streamoid Technologies
Bangalore, India
Led the development of Fashion AI platform with multiple groundbreaking projects over 10 years.
Key Projects:
- New Recommendation System: Hybrid (text+visual) search engine using CLIP and few-shot learning
- Allbirds Data Science: Predictive models for cart abandonment and shopper behavior using catboost
- Catalogix: Modern image editing APIs with BG removal, smart auto-resizing, shadow generation
- AI Studio: On-demand training and deployment of image classifiers for fine-grained fashion attributes
- Visual Search & Similar Products: Real-time feature extraction and visual product recommendations
- Autoscribe: Auto-scalable model pipeline for fashion attribute extraction
PyTorch CLIP Qdrant OpenCV FastAPI Catboost Google Cloud MongoDB Redis
Software Engineer
Aug - Dec 2013Samsung Research Institute (SRI)
Delhi, Noida, India
Developed prototype for real-time background removal in video calls using depth sensors.
Key Projects:
- Real-time video background removal system using depth sensor technology
Computer Vision Real-time Processing Depth Sensors
Education
M.S (Research) in Computer Vision and Machine Learning
2010 - 2013School of IT, IIT Delhi
New Delhi, India
Thesis: Multi-view Reconstruction using Relaxation Labeling
Proposed a novel approach for multi-view 3D reconstruction and compared it with state of the art algorithms.
B.Tech in Computer Science
2006 - 2010College of Engineering Roorkee
Roorkee, India
Bachelor Project: Offline Handwritten Devnagari Character Recognition
Tech used: MATLAB, LibSVM
Built with
Astrofolio
