2nd Place, ARM AI Developer Challenge
DreamMeridian runs a quantized xLAM-2-1B model on a Raspberry Pi 5 (ARM Cortex-A76) to answer natural language questions about geographic data with no cloud connectivity. Ask “Find hospitals near Camp 6” or “Show 15-minute walkable area from downtown Flushing” and get real routing results on a road network graph, not straight-line estimates. Pre-built datasets cover Cox’s Bazar, San Juan, and Jakarta. Responses in under 15 seconds on the Pi’s ARM NEON/dotprod pipeline.
Featured in NYU Tandon Engineering: “Meet Adam Munawar Rahman: Leveraging AI for Humanitarian Use” and covered in Business Weekly.