I am a Computer Science undergraduate at IIIT Sri City specializing in Artificial Intelligence and Machine Learning. My work spans distributed machine learning systems, AI infrastructure, control over lossy networks, and generative models for scientific computing.
I am the first author of research accepted at MICCAI 2026 (supported by a MICCAI Travel Award) and maintain several private research codebases currently under peer review. As a Google Summer of Code 2026 contributor with DeepChem, I integrated generative protein design architectures into the core library, and I actively contribute to Project HAMi for Kubernetes GPU resource virtualization.
| Project | Organization / Ecosystem | Description & Contributions |
|---|---|---|
| DeepChem | Google Summer of Code 2026 | Implemented RFDiffusion and RFDiffusion All-Atom generative protein design models into DeepChem's suite. |
| Project HAMi | Cloud Native Computing Foundation / K8s | Contributed to heterogeneous GPU resource sharing and device virtualization for Kubernetes clusters. |
| HAMi-core | Project HAMi Ecosystem | In-container GPU resource control and transparent hardware virtualization engine. |
- Right Answer, Wrong Evidence: Revealing Visual Confabulation in Multi-Image Medical VLMs
28th International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2026), Strasbourg, France
🏆 Recipient of the MICCAI 2026 Travel Award- Introduced the Evidence Attribution Score (EAS) to quantify whether multi-image medical vision-language models rely on genuine visual evidence vs. visual confabulation.
- Codebase:
eas-medical-vlm
The following research repositories contain active codebases for manuscripts currently undergoing double-blind peer review at premier IEEE journals and conferences. They are maintained privately and will be made public upon acceptance:
|
Multi-node GPU parallel training framework and communication bounds for parameter-efficient fine-tuning (LoRA, QLoRA). 🔒 Private Research Repository |
Information-theoretic bounds, error exponents, and min-cut graph routing algorithms for decentralized detection. 🔒 Private Research Repository |
|
Controlled set-invariance, anytime restoration algorithms, and Universal Zooming Quantizer simulators for expansive dynamical maps. 🔒 Private Research Repository |
Hyperspectral super-resolution architectures with physical mineral constraints for lunar remote sensing imagery. 🔒 Private Research Repository |
|
Unified state-space model for joint routing, traffic engineering, and congestion control validated via large-scale ns-3 simulations. 🔒 Private Research Repository |
Distributed PyTorch FSDP pipelines and graph neural network models for molecular docking and drug discovery (IIT Madras Research). 🔒 Private Research Repository |
Note: These repositories will be made public once the respective papers are accepted and published.
-
Google Summer of Code Contributor @ DeepChem (May 2026 – Aug 2026)
Integrated generative protein design pipelines (RFDiffusion & RFDiffusion All-Atom) with reusable APIs, unit tests, and benchmarking utilities. -
Data Science & AI Research Fellow @ IIT Madras (May 2026 – Jul 2026)
Engineered multi-GPU distributed PyTorch training pipelines (FSDP, mixed precision) for molecular binding affinity prediction using Graph Neural Networks and RDKit. -
Amazon ML Summer School 2026 (Jul 2026 – Aug 2026)
Selected for intensive training covering Large Language Models, Generative AI, Reinforcement Learning, and production ML infrastructure. -
AI Intern @ Infosys Springboard (Aug 2025 – Oct 2025)
Developed AI TraceFinder, a forensic document analysis system achieving 93.75% Scanner ID accuracy and 89.86% Tampered Document Detection.
Repository:AI_TraceFinder-• 🤗 Interactive Demo
|
Forensic document analysis and scanner identification using stacked ensemble models over 153 extracted forensic features.
93.75% Scanner ID • 89.86% Tampering Detection |
Multilingual healthcare assistant providing AI medication safety analysis and patient assistance.
🏆 Winner - IdeaVerse National Hackathon |
|
Multi-agent coffee shop assistant integrating LLM reasoning, RAG knowledge retrieval, and microservices. |
Multi-label retinal disease classification leveraging Vision Transformer architectures. |
|
Medical image restoration and noise reduction using Denoising Diffusion Probabilistic Models. |
Text-conditioned binary segmentation model for drywall construction quality analysis. |
┌───────────────────────────┬─────────────────────────────────────────────────────────────────┐
│ Category │ Technologies │
├───────────────────────────┼─────────────────────────────────────────────────────────────────┤
│ Core ML & AI │ PyTorch, Hugging Face, scikit-learn, OpenCV, RDKit, NetworkX │
│ Distributed & Systems │ CUDA, PyTorch DDP, FSDP, DeepSpeed ZeRO, SLURM, Docker, K8s │
│ Programming Languages │ Python, C/C++, Go, Java, SQL, JavaScript, TypeScript │
│ Cloud & Utilities │ Linux, Git, CMake, Neo4j, FastAPI, ns-3 │
└───────────────────────────┴─────────────────────────────────────────────────────────────────┘
📂 Early Foundational & Web Engineering Projects (Archived / Coursework)
The following repositories represent early coursework and foundational full-stack applications developed prior to specializing in distributed AI systems and scientific ML:
SkillLink/SkillLink_React: Full-stack home services and supplies platform.housekeeping-management-system: Web application for managing housekeeping operations.language-management-system: Database application for language learning tracking.newsletter-management-system: Full-stack web application for managing digital newsletters.Student-profile-management-system: Full-stack management app for student profiles.Programmer-survey: Professional programmer survey platform.