AxionOrbital Launches AI for All-Weather Satellite Imagery
AxionOrbital Space, part of Y Combinator's Winter 2026 batch, has launched its first foundation models for translating radar satellite data into human-readable optical imagery. The goal: making Earth observable around the clock regardless of weather or lighting conditions.
Traditional optical satellites are rendered useless roughly 70% of the time due to cloud cover and darkness. Synthetic Aperture Radar (SAR) satellites can see through clouds and operate at night, but the data they produce is unintelligible to humans and incompatible with standard computer vision pipelines.
AxionOrbital's models translate raw radar backscatter into analysis-ready optical imagery in real-time. The architecture uses deterministic one-step diffusion to transform radar signals into photorealistic images.
The technology targets three primary markets: defense, commodities trading, and disaster response.
The founding team brings relevant technical depth. CEO Dhenenjay Yadav was an ML Engineer at ISRO (the Indian Space Research Organisation) and RL Researcher at IIM Ahmedabad. CTO Atharva Peshkar is a computer vision researcher with experience at Harvard VCG and holds a CS PhD from CU Boulder.