India
I previously worked as a researcher at the Technical University of Munich (TUM), where I integrated AI with Molecular Dynamics software (LAMMPS) via MLMOD, testing AI models to enhance simulations in material science. During my Research at IIT Guwahati, I developed and evaluated metrics for AI-based data annotations, gaining valuable experience in optimizing AI performance. Beyond academics, I co-founded Treats, an IoT-based smart pet collar startup, and served as the Director of Societies for Olympism, the official sports club of Manipal University Jaipur, where I was recognized for my exemplary contributions. I bring technical expertise in AI/ML, IoT, and software development, with a passion for interdisciplinary research and innovation. In addition to my technical pursuits, I was a national-level basketball player and held the role of Creative Head for the Rotary Club, where I honed my leadership and teamwork skills.
• Built in-production proposals for Hopfield (Meta) and OSworld (NVIDIA and Alibaba) with the RnD team , DM GDPval • Built internal automation tooling for OSWorld and TerminalBench SFT data-collection pipelines, developing Pygame-based viewport interfaces and VM orchestration (VMware) to enable human annotators to record high-quality training trajectories for desktop AI agents. • Developed EvalGen AI, an agentic batch-review system that autonomously validates OSWorld task JSON configurations, reducing manual QA effort by 40% and improving deliverable turnaround time.
• Integrated LLM-driven predictive models and MLOps pipelines into Molecular Dynamics (MD) simulations using LAMMPS and MLMOD, reducing setup time by 30%. • Utilized CNNs and deep learning architectures for visual feature extraction in MD simulations, improving predictive accuracy by 20%. • Implemented RAG for on-demand model inference, deployed containerized solutions via Docker, ensuring reproducibility and streamlined CI/CD.
• Developed and evaluated metrics for AI-based data annotations, including BERTScore and MoverScore. • Integrated MENLI into annotation pipeline, outperforming other metrics and enhancing evaluation framework. • Collaborated on human verification tasks to ensure reliability of AI-generated annotations.
• Led a team to develop a secure, on-premise Llama 2 chatbot for the Indian Army using RAG, LoRA fine-tuning with 8-bit quantization, and a FastAPI + React.js stack with Docker and end-to-end encryption • Received official Appreciation Letter from Capt Shobhit Negi (Officer Commanding) for delivering a low-latency, policy-compliant chatbot handling confidential military data
• Organized orphanage drives and blood donation campaigns, increasing participation by 30% and enhancing community outreach. • Led community initiatives and events for 100+ members, fostering innovation, inclusion, and collaboration.