Adaptive Quantum-Enhanced Learning (NISQ Era)
Hybrid framework for robust QML; presented at the 5th Intl. Conf. on AI-ML-Systems.
CS undergrad at Symbiosis Institute of Technology. AI/ML researcher. Systems tinkerer building agentic AI, post-quantum pipelines, and generative medical vision.

I build things that sit at the seams — where quantum ML, agentic systems, and post-quantum crypto stop being buzzwords and start shipping. Four papers deep, five hackathons in, and currently exploring generative medical imaging.
Helped build a Flutter + Firebase adoption app with sign-in and cloud workflows.
Hybrid framework for robust QML; presented at the 5th Intl. Conf. on AI-ML-Systems.
Quantum-resilient multipath QUIC with learning-based routing and adaptive FEC.
Multi-agent RAG framework using the Indian Kanoon corpus for adversarial legal proceedings.
Sinkhorn OT regularizer prevents manifold collapse. +5.2% accuracy at low overhead.
High-fidelity medical modality translation using generative deep learning. This project translates MRI scans to CT scans (and vice versa) utilizing dual-generator GANs and DDPMs (Denoising Diffusion Probabilistic Models), reducing scan times and costs while preserving pixel-perfect anatomical structures.
An asynchronous secure file transfer pipeline built in Rust. It utilizes high-throughput LZ4 compression, post-quantum key encapsulation, and CRC32 payload verification to deliver lightweight, highly resilient secure data pipelines. The system yields +30% throughput, +40% network reliability, and 20–25% smaller secure payloads.
A LeetCode-style interactive preparation platform for USMLE (United States Medical Licensing Examination) exams. Powered by Pinecone vector databases and RAG, it serves over 1,000+ interactive clinical cases, adapting dynamically to user weak points and reducing exam prep time by approximately 35%.
An agentic AI courtroom simulator. Utilizing LangGraph for complex agent coordination, this system assigns autonomous judge and lawyer agent roles to analyze legal documents, automate argument extraction, cross-examine evidence, and draft briefs from 300+ civil lawsuit files.
An IoT-driven environmental risk analyzer. Built with hardware sensors (DHT11 temperature/humidity, MQ135 air quality) connected to a reinforcement learning model that analyzes real-time ambient hazards and predicts fire or air quality threats.