OCR
Guided model investment for “text-in-the-wild” on AR glasses, partnering with ML researchers to raise egocentric OCR accuracy from 26% to 39%. Defined target use cases and model gaps (text grouping and ordering, intent disambiguation) to prioritize investments.
Translate
Prototyped multimodal Live Translation on AR glasses, turning egocentric OCR and hand disambiguation research into a product experience.
Voice
Introduced Meta AI natural-language voice commands across first-party VR apps, defining a new hands-free interaction model for the surface.