Chethan Parameshwara

I am an AI researcher at Meta, where I train real-time multimodal LLMs and diffusion models.

Prior to joining Meta, I was a researcher at WaveForms AI, focusing on posttraining and distillation of multimodal diffusion models for audio-video generation. Earlier, as a researcher at Amazon, I was a core contributor to the Nova foundational models.

I did my PhD at the University of Maryland, College Park (UMD). My PhD research focused on video understanding/reasoning, 3D computer vision and neuroscience-inspired artificial intelligence.

Email  /  GitHub  /  Google Scholar  /  LinkedIn  /  X

My selected publications are listed here. The complete list of publications can be seen from my Google Scholar page.

Core contributor to Amazon Nova Reel, an advanced video generation model within the Amazon Nova foundation suite, designed to produce high-quality, customizable outputs with fine-grained motion control.

We describe a first step towards learning general-purpose visual representations of physical scenes using only image prediction as a training criterion.

We present a novel bipolar HD encoding mechanism designed for encoding spatio-temporal data of an asynchronous event camera.

We propose a novel differentiable programming technique for camera pose estimation in autonomous driving and flying scenarios.

We propose a bio-inspired neural network for the motion detection problem using an asynchronous event camera.

We present a novel approach to segment never-seen objects by physically interacting with them and observing them from different views.

We propose a hybrid solution to multi-object motion segmentation using a combination of model-based and deep learning approaches with minimal prior knowledge.

We present a deep learning-based solution for detecting propellers (to detect drones) for mid-air landing and following.

We present the first deep learning based solution for dodging multiple dynamic obstacles on a quadrotor with a single event camera and onboard computation.

We present a novel motion compensation approach for moving object tracking with an asynchronous event camera.

Meta
Aug, 2025 -- Present
Researcher

Real-time multimodal LLMs and diffusion models. Research on posttraining, distillation, quantization, and inference optimization.

WaveForms AI
May, 2025 -- Aug, 2025
Researcher

Research on real-time multimodal diffusion models for audio-video generation.

Amazon
Aug, 2022 -- May, 2025
Researcher

Research on video diffusion models (text-to-video, image-to-video, video-to-video) and 3D computer vision (NeRF, Gaussian Splatting, Neural Rendering, and SLAM).

University of Maryland, College Park (UMD)
Aug, 2017 -- Aug, 2022
Student Researcher

Research on video understanding (visual reasoning, object detection and segmentation, motion segmentation) and geometric computer vision (camera pose estimation, object motion estimation, neural rendering, SLAM)

SRI International (formerly Stanford Research Institute)
Summer, 2021
Visiting Researcher

Research on few-shot deep learning and gradient-free learning algorithms.

Thanks to Jon Barron for this minimalist website template.