Chinmay Talegaonkar

I am a final-year Ph.D. student at UCSD, working broadly in 3D vision and computational photography, with a particular focus on leveraging diffusion models to solve problems across these domains. . I am advised by Prof. Nicholas Antipa, and have worked closely with Yash Belhe and Prof. Ravi Ramamoorthi. I completed my M.S. in ECE from UCLA, advised by Prof. Achuta Kadambi and B.Tech in EE with a minor in CS from IIT Bombay.

Prior to joining UCSD, I was senior deep learning engineer at Akasha Imaging (acquired by Intrinsic.ai, now part of Google). At Akasha, I worked on developing a Multiview 6DOF Pose estimation system for industrial automation. I also played a lot of Ping Pong with Agastya Kalra.

Internships:

Meta — Researching 3D generative models for scene generation with Tom Monnier and Andrea Vedaldi. Summer and Fall 2026.
Amazon Lab126 — Training a VLM for home layout estimation from 3D point clouds. Summer 2025.
Qualcomm AI — Developing memory-efficient methods for dynamic human reconstruction, resulting in a US Patent. Summer 2023.
NVIDIA — Developing and optimizing CUDA kernels, with contributions that are now part of CUTLASS and NVIDIA RAPIDS. Summers 2018 and 2020.

I am on the industry job market for roles in 2027. I am looking for research internships starting January 2027, and full-time roles starting March or June 2027. Please reach out if you think I'd be a good fit!

Email / CV / Short CV / Scholar / Twitter / LinkedIn / Github

Research

In a nutshell, I enjoy generating pixels, voxels, worlds (and world models). My Ph.D. research focuses on coupling the physics of image formation with diffusion models (image/video) and 3D representations (NeRFs, Gaussian Splatting) to solve inverse recovery problems in imaging. I like working at the boundary of perception, reconstruction, and generative modeling of the visual world. Excited to explore world modeling and robotics through this lens.

	Single-Shot HDR Recovery via a Video Diffusion Prior Chinmay Talegaonkar, Jinshi He, Nicholas Antipa NeurIPS 2026 arXiv / Code Release
	Repurposing Marigold for Zero-Shot Metric Depth Estimation via Defocus Blur Cues Chinmay Talegaonkar, Nikhil Gandudi Suresh, Zachary Novack, Yash Belhe, Priyanka Nagasamudra, Nicholas Antipa NeurIPS 2025 (Spotlight) Project Page / Code Release /
	Volumetrically Consistent 3D Gaussian Rasterization Chinmay Talegaonkar, Yash Belhe, Ravi Ramamoorthi, Nicholas Antipa CVPR 2025 (Highlight) Project Page / Code Release /
	RnGCam: High-speed video from rolling & global shutter measurements Kevin Tandi, Xiang Dai, Chinmay Talegaonkar, Gal Mishne, Nicholas Antipa ICCV, 2025
	Pose Estimation of Buried Deep-Sea Objects with 3D Vision Deep Learning Models Jerry Yan, Chinmay Talegaonkar, Nicholas Antipa, Eric Terrill, Sophia Merrifield OCEANS, 2024
	Visual Physics: Discovering Physical Laws from Videos Pradyumna Chari, Chinmay Talegaonkar, Yunhao Ba, Achuta Kadambi arXiv Preprint, ICCP 2020 Poster* ICCP Poster Video A revised version of this paper was published in IEEE Access .
	Compressive Phase Retrieval under Poisson Noise Chinmay Talegaonkar, Parthasarathi Khirwadkar, Ajit Rajwade ICIP, 2019
	Performance Bounds for Tractable Poisson Denoisers with Principled Parameter Tuning Chinmay Talegaonkar, Ajit Rajwade GlobalSIP, 2018

Miscellanea

Awards & Fellowships

GuruKrupa Fellowship at UCLA
South East Asia ML Summer School (SEAMLS) -- top 100 out of 1100 applicants.
Silver medalist at IJSO 2013, representing the Indian team.

Thanks Jon Barron for the website source code.