On-Device Multimodal Perception Call for Proposals - Fall 2026
On-Device Multimodal Perception Call for Proposals - Fall 2026 supports researchers in visual generation, encoding, and sensor fusion to turn limited device observations into useful experiences and reliable tracking.
Funder: Amazon Research Awards
Due Dates: October 1, 2026 (Submissions open) | November 4, 2026, 11:59 p.m. Pacific Time (Submission deadline)
Funding Amounts: One-year awards: cash no more than $80,000 USD on average and AWS Promotional Credits no more than $40,000 USD on average
Summary: Supports on-device visual generation, efficient visual encoding, and multimodal sensor fusion that turn limited device observations into useful experiences and reliable tracking.
Key Information: Each principal investigator may submit only one proposal to ARA per call period.
Description
Amazon Research Awards, through Amazon’s Devices and Services organization, supports research that closes the gap between what compact devices capture and what users need. The call welcomes theoretical advances, creative ideas, and practical applications using learned on-device processing or generative methods. Research should address noisy, incomplete, or sparse observations within fixed power, bandwidth, and thermal constraints.
Proposals may address multiple topics: generating intended visual experiences from missed moments, poor viewpoints, or discontinuous recordings; efficiently encoding visual inputs for on-device vision-language models; and learning to fuse inertial, visual, and ranging streams for object tracking. Priorities include evaluation beyond standard reconstruction benchmarks, balancing encoding cost with relevance and open-ended queries, calibration-free deployment, and continued tracking during extended sensor occlusion.
