I'm an MS student in Computer Science at UMass Amherst with a 4.0 GPA, focused on the intersection of machine learning, natural language processing, and computer vision. My research spans multi-agent reinforcement learning, deep learning for side-channel security, and AI-assisted data mining.
Currently a Research Extern at Amazon Web Services, where I work on multi-agent RL fine-tuning for LLMs in game-theoretic settings. Previously, I interned at the Indian Academy of Sciences (semiconductor ML) and Widezo (voice AI platforms).
I hold a BE in Computer Science from Anna University, CEG Campus with a 9.20 GPA. Interested in research collaborations or opportunities? Let's connect.
- Researched multi-agent RL fine-tuning for LLMs in non-cooperative game-theoretic settings
- Implemented DPO and IPO post-training pipeline using DeepSeek-R1-7B and Qwen2.5-7B across 30+ games
- Designed AI vs. AI game-playing system generating 500+ preference-labeled examples, improving strategic performance by 38%
- Improved sampling efficiency of replay buffers in reinforcement learning algorithms using statistical mathematics
- Implemented Successor Feature (SF) learning in grid-world and four-rooms environments
- Investigated four sampling strategies including a novel norm-based approach for SF convergence
- Built full-stack voice AI platform using LangChain, CrewAI and React with real-time voice recognition and RAG
- Enabled deployment of customizable voicebots across 5+ use cases including customer service and sales automation
- Optimizing Helmet Detection With Hybrid YOLO Pipelines: A Detailed AnalysisV Muthuraman, D Dharma, V Chandrasekaran, B Senthil VelanCS & IT-CSCP 14 (24), 83–933 citations
- An Ensemble MLP-RF Model For The Prediction Of DG-MOSFETs: Addressing Fabrication Process VariationsV Muthuraman, K Sehra, V Kumari, M SaxenaVLSI Design 2025 / IEEE1 citations
- KeystrokeNET: Deep Learning for Side-channel AttacksV Muthuraman et al.Preprint 2025
- Research interests: multi-agent RL, LLM alignment, computer vision for accessibility
- Open to research collaborations and PhD opportunities in AI/ML
- 4 total citations on Google Scholar (h-index: 1)
Feel free to reach out for research collaborations, questions, or opportunities.