On-Device Multimodal Perception call for proposals - Fall 2026

Support research on generative visual experiences from incomplete capture, efficient visual encoding for on-device vision-language models, and learned multi-sensor object tracking. Proposals should explain operating constraints, scientific novelty, evaluation trade-offs, a feasible one-year plan, open-source outputs and intended AWS tools. Use the ARA template, normally up to four pages excluding appendices. The submission period is 1 October–4 November 2026, closing at 23:59 Pacific Time; decisions are expected in February 2027. Eligible faculty and permanent nonprofit researchers apply through Amazon’s research portal, with one primary-PI proposal per call period.

Full-time faculty at an accredited academic institution or permanent researchers at a legally recognised non-governmental nonprofit. One primary-PI proposal per call period; co-PIs must also qualify. Amazon employment/contracting in the previous 12 months, specified household conflicts and Amazon cloud-procurement authority disqualify applicants. Government employees are excluded except university employees. Applicants must meet age-of-majority and programme rules.