I'm a third-year undergrad interested in computer vision, generative models, and image processing. My research focuses on understanding and improving how visual information is represented, generated, and reconstructed by learning-based systems. I have worked on problems spanning generative image watermarking, biomedical imaging, and inverse imaging. I enjoy exploring ideas at the intersection of deep learning, signal processing, and visual computing.
Looking for PhD opportunities in computer vision and robotics domain.
News
Beyond Plausible Explanations accepted at IEEE Tencon 2026
Education
Indian Institute of Information Technology, Jabalpur
B.Tech in Electronics and Communication Engineering
Publications
ECCV 2026
BiSLW: Bi-Spectral Latent Watermarking for Generative Diffusion Models
Aryan Pandit
Designed a dual-frequency latent watermarking framework that independently embeds information into low- and high-frequency latent components, improving visual fidelity while maintaining robust watermark recovery under regeneration attacks and common image distortions.
IEEE Region 10 TENCON 2026
Beyond Plausible Explanations: Verification-Aware Reasoning for Reliable LLM Fact Checking
Aryan Pandit
Introduced a verification-aware reasoning framework integrating evidence verification with structured reasoning for reliable LLM-based fact checking.
IEEE SCEECS 2026
🏆 Best Paper Award
Plug-and-Play Proximal Graphical Model for Demosaicing of Microscopy Images
Aryan Pandit, Anurodh Kumar, Amit Vishwakarma
Plug-and-Play proximal optimization framework integrating BM3D priors for high-resolution microscopy image demosaicing. Achieved an average PSNR of 41.06 dB and SSIM of 0.909 on 1,324 microscopy images, demonstrating superior reconstruction quality over existing methods.
Experience
Research Intern — Indian Institute of Information Technology, Jabalpur
Advisor: Dr. Amit Vishwakarma
Co-Authored a research paper proposing a Plug-and-Play Proximal Graphical Model using BM3D priors for robust demosaicing of high-resolution microscopy images. Achieved an average PSNR of 41.06 dB and SSIM of 0.909 on 1324 high-res images, surpassing previous baselines.
Open-Source Developer — Iabsis HCW - C4GT Program
Organizer: Code For Government Tech
Developed and integrated RBAC system and configured messaging services (SMS and WhatsApp) to enhance the HCW admin platform’s security and communication capabilities. Worked with NestJS for backend services and Angular for frontend integration, focusing on modular, scalable system design.
Projects
TARS
Cross-platform desktop app that automates SEO audits, content updates, performance checks, and error fixes using Gemini LLM with a custom RAG pipeline. Reduced manual SEO effort by 70% and improved Lighthouse scores by up to 20% through parallel crawling of 100+ pages and semantic retrieval with caching optimization.
STASH
Windows session manager that captures and restores multi-application workspaces in under 2 seconds, saving 30–60 minutes of daily setup time. Built using a hybrid Rust–Python architecture achieving 2x faster state restoration with 99% session recovery accuracy across 100+ test cycles.
Tipsy
AI-powered Chrome extension that analyzes JavaScript runtime error types and suggests real-time fixes using Gemini API and Chrome DevTools Protocol. Reduced debugging time by 40% with sub-2 second response latency and lightweight injection overhead (<100ms) to maintain performance.