Xander Backus

Xander Backus is affiliated with the MIT Media Lab and studied at MIT from 2022 to 2026. His listed coursework includes machine learning, multimodal AI, natural language processing, algorithms, and robotics. He collaborated with me on EgoBlind-RA, a project on risk-adaptive multimodal AI for blind and low-vision users.

View Xander’s LinkedIn profile · Read about EgoBlind-RA

EgoBlind-RA: Towards Safer Egocentric Assistive AI for Blind Users via Risk-Adaptive Routing
EgoBlind-RA: Towards Safer Egocentric Assistive AI for Blind Users via Risk-Adaptive Routing

EgoBlind-RA routes egocentric queries from blind and low-vision users by urgency using a lightweight CLIP classifier (0.905 ROC-AUC), then allocates response policies matched to their safety stakes — showing that the value of DPO and the cost of routing noise both depend sharply on architectural regime.

May 12, 2026