Skip to main content

Anirudh Mamgain

Artificial Intelligence • Computer Vision • Vision-Language Models • Reinforcement Learning

Email | LinkedIn | GitHub

About Me

My research lies at the intersection of Artificial Intelligence, Machine Learning, Computer Vision, Vision-Language Models, and Reinforcement Learning. I am particularly interested in developing intelligent systems that combine perception, reasoning, and decision-making to solve real-world problems. My recent work spans visual reasoning, multimodal AI, autonomous perception, and intelligent surveillance.
 

Master's Thesis

Gains and Trade-offs of Reinforcement Learning for Reasoning in Vision-Language Models

This research investigates how reinforcement learning can improve reasoning capabilities in Vision-Language Models (VLMs). The work explores modern post-training techniques, including Group Relative Policy Optimization (GRPO), to enhance multimodal reasoning while analyzing the trade-offs between reasoning performance, computational efficiency, and model generalization. The study evaluates various prompting strategies, reward designs, and fine-tuning methods across visual reasoning benchmarks to better understand their impact on accuracy and inference efficiency.

Relevant Coursework

  • Machine Learning Principles and Techniques
  • Deep Learning: Theory and Practice
  • Simulation Modeling and Analysis
  • Engineering Statistics
  • Probability and Stochastic Processes
  • Mathematical Optimization Techniques
  • Applied Integer Programming
  • Quantitative Models for Supply Chain Management
Anirudh Mamgain