Maitreya Suin

Logo

Senior Researcher, AI

Publications

Notes

LinkedIn
Google-Scholar
E-Mail
CV

I am working as a Senior Researcher, AI at Samsung AI Center Toronto. Before that, I was a postdoctoral researcher in the AIEM Lab, Johns Hopkins University. I was advised by Prof. Rama Chellappa specializing in topics related to Computer Vision and Machine Learning. Previously, I completed my MS and PhD at Image Processing and Computer Vision Lab, IIT Madras under the supervision of Prof. A. N. Rajagopalan.

My research expertise lies in the domains of computer vision, machine learning, and multimodal generative AI, with a specific focus on diffusion models, vision-language models, post-training, camera image-processing pipelines, image and video restoration, and image/face recognition and detection. My recent work involves leveraging foundation models, including diffusion models and VLM/LLM-based reasoning systems, for tasks such as degradation-aware image-quality analysis, camera-pipeline triage, blind face restoration, and few-shot personalized generation/editing, with an emphasis on perceptual quality, identity preservation, reliability, and efficiency. During my PhD, I focused on building adaptive neural networks for handling spatially and temporally varying tasks, using efficient non-local attention and spatially varying filtering modules for images and videos.

Writing

I write Modern AI, Under the Hood — deep technical notes on the modern AI stack, with runnable code, derivations, and measured benchmarks.

Recent: The KV Cache: Build It, Break It, Measure It · FlashAttention: The Matrix You Never Have to Write Down · A Tensor Is a Flat Buffer and a Promise

Updates