I am a B.Tech Information Technology student at NITK Surathkal (CGPA: 8.55) specializing in Computer Vision, Domain Adaptation, LLM Fine-tuning & Alignment, and Full-Stack AI Web Applications.

• 🔬 Research Assistant @ Xu Lab, Carnegie Mellon University (CMU) — Few-Shot Cryo-ET subtomogram classification & Hierarchical Domain Adaptation.
• 🧪 Research Intern @ VIDA Lab, NTU Singapore — Dataset curation & LLM instruction synthesis pipelines for text-to-infographic generation.
• 🤖 CV & LLM Research Intern @ HALE Lab, NITK — Fine-tuning 9B & 35B parameter LLMs using DPO, GRPO, and multi-GPU VRAM optimization.
• 🏢 ML Developer @ IRIS NITK — Production ML models (Sentiment Analysis, LLaMA-7B placement summarizer) for 7,000+ staff and staff.

• Carnegie Mellon University (Xu Lab) | Researcher (May 2026 – Present)

  • Developed Hierarchical Domain Adaptation frameworks for few-shot Cryo-ET subtomogram classification on Swin3D backbones.
  • Achieved +6.05% gain in 3-shot classification accuracy across low-shot target datasets (Noble & Qiang).

• Nanyang Technological University (VIDA Lab) | Research Intern (Jul 2025 – Present)

  • Constructed aligned image-structured instruction pairs for infographic diffusion pretraining (published on Kaggle, cited in InfoAffect 2025).
  • Synthesized 50 papers across IEEE VIS, TVCG, and ACM CHI for automated infographic generation strategies.

• NITK Surathkal (HALE Lab) | Computer Vision & LLM Research Intern (May 2026 – Jul 2026)

  • Fine-tuned Ornith 9B & 35B parameter LLMs, improving downstream task accuracy by 24% and cutting alignment compute costs by 50% (DPO/GRPO).
  • Reduced peak VRAM consumption by 40% across 100+ evaluation pipelines via custom 2D tensor auto-scaling.

• Mohamed bin Zayed University of AI (MBZUAI) | Summer Research Intern (May 2025 – Jul 2025)

  • Designed semi-supervised deep clustering for 3D Cryo-ET tomograms combining YOPO feature extraction with GMMs.

• IRIS, NITK | Machine Learning Developer (Sep 2024 – Present)

  • Deployed production ML modules into university ERP serving 7,000+ staff and staff.

• Full-Stack AI Developer Platform — Serverless Next.js 16 + React 19 portfolio with interactive CLI shell and sub-50ms page transitions.
• Hierarchical Cryo-ET Domain Adaptation — Few-shot subtomogram classification framework using Swin3D, STN, MMD, and CORAL losses.
• Twinscribe / Infographics LLM & Diffusion Pipeline — Automated step-wise instruction synthesis and dataset curation for infographic generation.
• Vision Kinect — Hands-free Tetris game driven by real-time gesture recognition using YOLOv5 and OpenCV (<15ms latency).
• Imagined Risk — Geometric world models and multi-view vision-language RL for autonomous driving perception.
• 3D CNN Video Classification — Spatio-temporal action recognition engine in PyTorch.

• Languages: Python, C++, C, JavaScript, TypeScript, Java, HTML/CSS
• Full-Stack & Web: Next.js 16, React.js, Tailwind CSS, Node.js, FastAPI, REST/Serverless APIs, tRPC
• AI / ML / CV: PyTorch, TensorFlow, LLM Alignment (DPO, GRPO), Stable Diffusion, Swin3D, YOLOv5, OpenCV, HuggingFace
• Tools & Infra: Git, Docker, Linux, MLflow, EMAN2, CryoSPARC, Kaggle, LaTeX

_{Connect with me via LinkedIn or explore my Live Portfolio Website.}