Open to opportunities

Vivek Narula

@viveknarula

Message

I build computer vision and generative AI systems for real-time video analytics and autonomous workflows.

India

Message

What I'm looking for

I'm looking to build and deploy practical AI systems, especially in computer vision, multimodal deep learning, generative AI, autonomous agents, and RAG. I value opportunities to turn research into efficient, real-world products with cross-functional teams.

I'm building real-time video analytics and generative AI workflows, translating multimodal deep learning research into deployable systems.

In my M.Tech research, I engineered a slot-based attention model for weakly supervised video anomaly detection that achieved state-of-the-art performance on UCF-Crime and XD-Violence. The work reduces annotation costs by using video-level labels and has been submitted to Computer Vision and Image Understanding.

At DTU's MLR Lab, I researched human action recognition and benchmarked VideoMAE, Swin, and two-stream CNN architectures for real-time video perception. I've also published two 2026 IEEE CICN papers on weakly supervised video anomaly detection.

I'm currently a Data Science Intern at PharmEasy, applying machine learning to predictive modeling and business pipelines. My hands-on stack includes PyTorch, OpenCV, RAG, LangChain, LangGraph, LangSmith, vision-language models, and LoRA.

Experience

Work history, roles, and key accomplishments

Current

Data Science Intern

Current

PharmEasy

May 2026 - Present (4 months)

Applied data science and machine learning techniques to optimize predictive modeling and business pipelines. Developed and fine-tuned data-driven algorithms, collaborating with cross-functional teams to deploy robust analytical solutions.

Data Science Machine Learning predictive modeling Python Collaboration

ML Research Intern

MLR Lab, Delhi Technological University

Jun 2024 - Aug 2024 (2 months)

Spearheaded research in Human Action Recognition, evaluating models from classical two-stream CNNs to modern Transformer-based architectures. Benchmarked inference latency and accuracy of SOTA models to identify optimal architectures for real-time video perception.

VideoMAE CNN Transformers Benchmarking PyTorch