ML Research Engineer at Snap Inc. | Building the next generation of Multimodal AI & Generative Models | Deep Learning, Computer Vision & Diffusion Transformers
I'm a passionate ML researcher and engineer with a strong foundation in electrical engineering and computer science (IIT Kharagpur). Currently at Snap Inc., I specialize in advancing multimodal AI systems and generative models. I'm driven by solving complex problems at the intersection of vision, language, and generative modeling.
- 🔬 Current Focus: Vision-Language Models (VLMs), Diffusion Transformers, and Multimodal System Training
- 🚀 Active Projects: Fine-tuning Diffusion Models, Model Optimization, Scalable AI Deployment Pipelines
- 📚 Learning: Virtual Try-On (VTON) Systems, Advanced Multimodal Architectures
- 🎓 Background: Dual degree in Electrical Engineering & Computer Science from IIT Kharagpur
- 💼 Experience: Snap Inc. (ML Research Engineer) | NeuroPixel.AI Labs | Analog Devices
Languages & Core Skills
- Python, C++, CUDA Deep Learning & AI
- PyTorch, OpenCV, FastAPI
- Diffusion Models, Vision Transformers, Vision-Language Models
Production & Optimization
- TensorRT (Model Optimization)
- Docker, Kubernetes, Git
- Scalable Deployment Pipelines
Data & Infrastructure
- Kafka, Distributed Systems
- Large-scale Model Training
Generative AI Research | Velocity Field Prediction with ViT
- Developed acceleration-based flow matching framework incorporating second-order trajectory dynamics
- Implemented velocity field prediction on Vision Transformers for efficient generative modeling
- Advanced flow-matching research enabling faster, more stable diffusion-based generation
- Eliminates curvature-induced numerical errors in trajectory paths for improved few-step sampling
- Tech: PyTorch, Vision Transformers, Flow Matching, Generative Models, VAE
- Built
Eklavya 7.0, a self-driving robot capable of autonomous navigation and obstacle avoidance. - Developed motion control, tuned EKF-based localization, and designed the robot's electronic architecture.
- Designed Multilevel Feedback Queue Scheduling in PintOS to enhance performance fairness.
- Developed a multiple producer-consumer system using shared memory to avoid deadlocks.
- 🔬 Contribute to foundational AI research in multimodal systems and generative models
- ⚡ Optimize real-world AI applications for production-scale deployment
- 👥 Mentor aspiring ML engineers and foster knowledge sharing
- 📖 Publish research and technical insights on cutting-edge AI techniques
I'm always interested in discussing:
- 🤖 Multimodal AI & Generative Models
- 🎨 Computer Vision & Image Generation
- 🚀 Scalable AI Systems & Optimization
- 🔍 Research Collaborations
Reach out via:
Thanks for visiting! Feel free to explore my projects and get in touch. Let's build something amazing together! 🚀