I am a Master’s student in Data Science and Artificial Intelligence at Saarland University. Previously, I was a part of the Smart Service Engineering (SSE) group at DFKI Saarbrücken, where I contributed to energy-efficient AI research within the ESCADE project.

My master’s thesis was on LLM-Guided Reinforcement Learning in Sparse-Reward Environments, investigating how schema-constrained LLM guidance can improve exploration in sparse-reward environments. My research interests are in Reinforcement Learning, Agentic Planning, and Foundation Models for RL.

I intend to graduate in early 2026 and am actively seeking opportunities in RL, LLM-guided planning research, and applied AI/ML roles.

Projects

• Membership Inference Attack on Image Classification Models
  Implemented a Membership Inference Attack (MIA) pipeline for CIFAR-10 and TinyImageNet classifiers using shadow models to evaluate training data leakage.
• Serverless Image Captioning API
  Engineered a serverless inference service with optimized cold-start management and early input rejection. Implemented failure-aware CI/CD pipelines to validate execution under strict constraints.

Publications

• LLM-Guided Reinforcement Learning in Sparse-Reward Environments (Preprint)
  Jain, V., Grossmann, G.
  arXiv preprint, 2025
• ESCADE: Energy-efficient Artificial Intelligence for Cost-effective and Sustainable Data Centers (Workshop)
  Janzen, S., Stein, H., Trinley, K., Agnes, C., Jain, V., Rajshekar, K., Shenoy, N., Rusch, A., Ghosh, S., & Maaß, W.
  Research Projects Exhibition at the International Conference on Advanced Information Systems Engineering (CAiSE), 2025
• Neuromorphic hardware for sustainable AI data centers (Peer-Reviewed)
  Vogginger, B., Rostami, A., Jain, V., Arfa, S., Hantsch, A., Kappel, D., … & Maaß, W.
  Neuro Inspired Computational Elements Conference (NICE), 2024
• Understanding interviewees’ perceptions and behaviour towards verbally and non-verbally expressive virtual interviewing agents (Peer-Reviewed)
  Thakkar, J. H., Rao, P. S. B., Shubham, K., Jain, V., & Jayagopi, D. B.
  Companion Publication of the International Conference on Multimodal Interaction (ICMI), 2022
• An intelligent model based on integrated inverse document frequency and multinomial Naive Bayes for current affairs news categorisation (Peer-Reviewed)
  Kumar, S., Sharma, A., Reddy, B.K., Sachan, S., Jain, V., Singh, J.
  International Journal of System Assurance Engineering and Management, 2022