Pixels are the new text.
Billions of satellite and aerial images sit unsearched and unindexed – the largest dataset on the planet, almost none of it queryable. Foundation models change that.
A letter from the founders
The Earth should be as searchable as the web.
Every road, refinery, forest, port, farm and city block on the planet is constantly observed by satellites, aircrafts and sensors. The data exists. The intelligence does not. Asking a simple question today – How many cargo ships are queued at the world's busiest ports right now? Which data centers were built this quarter? Which neighborhoods flooded today? – takes specialized analysts, a stack of disconnected tools and a lag of 48 hours. It should take one query and 60 seconds.
TerraByte is building the foundation models and secure mission platform that make the Earth queryable across decades of archive imagery, across every modality of Earth observation and increasingly in real time on the sensors that capture it. Imagine, if at the touch of a button, big tech companies could scout the perfect site for their next data center, city officials could mount wildfire rescue-efforts with great precision, or journalists could map out thinning icecaps in the Arctic.
The way Google indexed the web, we are indexing the physical world.
What we believe
A model is just the start.
Anyone can train one. The hard part and the real work is turning a question into a decision you can act on. That's what we build.
AI should run on the satellite, not just on the ground.
More satellites today carry onboard compute. Soon they'll carry GPUs. Eventually, entire data centers will orbit the Earth. TerraByte is built to run anywhere, from decades of archive imagery on the ground to a camera capturing a picture right now in orbit, on every kind of sensor and signal that watches Earth – so decisions happen in seconds, not hours.
Earth intelligence belongs to everyone.
Analysts, climate scientists, journalists and city planners deserve a common substrate.
Our models observe the Earth's features – ports, refineries, forests and cities – not faces or individuals. The planet is observable, the people in it remain unseen.
— Rishi Madhok & Fuxun Yu Co-founders. Previously at Microsoft - Planetary Computer.
The team
Who's building this.
Rishi Madhok
CEO & Co-founder
Rishi Madhok is a geospatial AI and computer vision leader with deep experience across remote sensing and applied ML. He has authored 18+ papers and patents in top AI/CV venues, including NeurIPS, ICLR, ICML, and CVPR. He earned a Master's degree with a focus on computer vision from Carnegie Mellon University. He started his career building perception systems for self-driving vehicles at Uber ATG and later joined Microsoft, where he served as a Principal Applied Science Manager and led geospatial AI initiatives for Microsoft Planetary Computer.
Fuxun Yu
CTO & Co-founder
Dr. Fuxun Yu has extensive expertise in geospatial AI, remote sensing, and computer vision. He has published 50+ academic papers in top AI, CV, and ML conferences including NeurIPS, ICLR, and ICML, with 1900+ citations. He graduated with his PhD from George Mason University in 2022. After graduation, he joined Microsoft as Principal Research Manager and led the Geospatial Foundational Model (GFM) project at Microsoft.
Zain Munad
Founding Engineer
Zain Munad is a founding engineer. He cut his teeth on NLP problems by applying LLMs at Kaiser Permanente, then worked on AI agents that attack MCP servers to find vulnerabilities. He is a UC Davis alum.