New Way Now: Improving connectivity across space with Satlyt
지금 보고 있는 영상
New Way Now: Improving connectivity across space with Satlyt
조회수 565
영상 설명 · 눌러서 펼치기
*Featured in this video:* Rama Ochieng Afullo, Founder & CEO at Satlyt
*Executive summary:* As the space economy rapidly expands, satellite operators face data bottlenecks and high downlinking costs. Decentralized computing provider, Satlyt, adopted Google Cloud to build an intelligent, in-orbit data processing network to improve satellite efficiency and save on these costs. By using Google Kubernetes Engine (GKE) for simulation and deploying models like Gemma to the edge, Satlyt transforms isolated satellites into an in-orbit AI data center. This autonomous system processes data in situ, and the company is targeting to reduce communication latency to under 20 milliseconds and make satellites up to 45% more power efficient. As it scales, Satlyt aims to save providers up to 90% on downlinking costs and help make space communications as seamless as accessing the internet.
*Challenge:* Despite pushing the frontiers of human exploration, much of the space industry’s infrastructure is outdated. The technology required for downlinking — sending data from satellites to Earth — is the third largest cost center for satellite operators due to a wide range of environmental and technical limitations. Operators frequently spend hundreds of thousands of dollars to downlink terabytes of images, only to discover that 90% of the data is completely corrupted or useless. Satlyt is using its technology to help these providers by preparing the space economy with in-orbit, intelligent data processing, and that requires low-latency, highly accurate, and efficient AI tools.
*Solution:* Satlyt built its software-defined network on Google Cloud to set the foundation for a virtual, AI data center in space. Before launching into orbit, Satlyt uses Gemini and Google Cloud as its testing and training ground. With Google Kubernetes Engine (GKE), the team runs complex space simulations to validate their architecture. The company also deploys small language models like Gemma directly onto satellites to run AI inference on the edge. This in-orbit compute allows satellites to autonomously cross-link, diagnose faults, and monitor health without waiting for ground instructions.
*Results:* By processing data in situ and using natural language tasking to daisy-chain satellite systems together, Satlyt eliminates data bottlenecks and creates a standardized communications protocol for the future of space. This in-orbit processing, combined with improved power efficiency helps reduce the volume of data sent back to Earth.
Running models like Gemma and Gemini 2.5 Flash helps Satlyt analyze system logs and minimize the volume of useless data sent back to Earth. The company is targeting a 45% improvement in power efficiency by making these changes. Satlyt is also working to use a standardized communication protocol to federate processing, targeting communication latency below 20 milliseconds and enabling near real-time interoperability. All together, the company is aiming to save providers up to 90% on costs associated with in-orbit data processing. By transforming isolated hardware into a federated network of intelligent nodes, Satlyt is also working to make talking to a satellite as easy as accessing the internet, opening the space economy to anyone.
*Key takeaways and highlights from our interview with Rama Ochieng Afullo, Founder & CEO at Satlyt:*
→ “Google really is our training ground for everything that we do in space. We first deploy activity on-prem and simultaneously deploy it on Google Cloud where we can test, train and figure out different strategies for what we would test in space.”
→ “ If we can operate on the edge with small language models like Gemma, we can reduce the amount of data teams send during downlinking by compressing it and federating satellite systems. That change can help us save satellite operators hundreds of thousands of dollars and also stack up other managed services from other software vendors to create a true marketplace in space.”
→ “With Google Cloud, Satlyt is creating a new way to scale compute beyond the limits of our planet.”
*Google Cloud products used:* Google Cloud, Google Kubernetes Engine, Gemini, Gemma, Gemini Flash 2.5, Google Workspace
Learn more:
→ Satlyt - Onboard AI & In-Orbit Computing https://satlyt.ai/
→ How Satlyt is building local AI for space with Gemma https://deepmind.google/models/gemma/gemmaverse/satlyt/
→ Make work less work with Gemini https://cloud.google.com/ai
→ Kubernetes, evolved: The foundation for platform builders https://cloud.google.com/kubernetes-engine
*Executive summary:* As the space economy rapidly expands, satellite operators face data bottlenecks and high downlinking costs. Decentralized computing provider, Satlyt, adopted Google Cloud to build an intelligent, in-orbit data processing network to improve satellite efficiency and save on these costs. By using Google Kubernetes Engine (GKE) for simulation and deploying models like Gemma to the edge, Satlyt transforms isolated satellites into an in-orbit AI data center. This autonomous system processes data in situ, and the company is targeting to reduce communication latency to under 20 milliseconds and make satellites up to 45% more power efficient. As it scales, Satlyt aims to save providers up to 90% on downlinking costs and help make space communications as seamless as accessing the internet.
*Challenge:* Despite pushing the frontiers of human exploration, much of the space industry’s infrastructure is outdated. The technology required for downlinking — sending data from satellites to Earth — is the third largest cost center for satellite operators due to a wide range of environmental and technical limitations. Operators frequently spend hundreds of thousands of dollars to downlink terabytes of images, only to discover that 90% of the data is completely corrupted or useless. Satlyt is using its technology to help these providers by preparing the space economy with in-orbit, intelligent data processing, and that requires low-latency, highly accurate, and efficient AI tools.
*Solution:* Satlyt built its software-defined network on Google Cloud to set the foundation for a virtual, AI data center in space. Before launching into orbit, Satlyt uses Gemini and Google Cloud as its testing and training ground. With Google Kubernetes Engine (GKE), the team runs complex space simulations to validate their architecture. The company also deploys small language models like Gemma directly onto satellites to run AI inference on the edge. This in-orbit compute allows satellites to autonomously cross-link, diagnose faults, and monitor health without waiting for ground instructions.
*Results:* By processing data in situ and using natural language tasking to daisy-chain satellite systems together, Satlyt eliminates data bottlenecks and creates a standardized communications protocol for the future of space. This in-orbit processing, combined with improved power efficiency helps reduce the volume of data sent back to Earth.
Running models like Gemma and Gemini 2.5 Flash helps Satlyt analyze system logs and minimize the volume of useless data sent back to Earth. The company is targeting a 45% improvement in power efficiency by making these changes. Satlyt is also working to use a standardized communication protocol to federate processing, targeting communication latency below 20 milliseconds and enabling near real-time interoperability. All together, the company is aiming to save providers up to 90% on costs associated with in-orbit data processing. By transforming isolated hardware into a federated network of intelligent nodes, Satlyt is also working to make talking to a satellite as easy as accessing the internet, opening the space economy to anyone.
*Key takeaways and highlights from our interview with Rama Ochieng Afullo, Founder & CEO at Satlyt:*
→ “Google really is our training ground for everything that we do in space. We first deploy activity on-prem and simultaneously deploy it on Google Cloud where we can test, train and figure out different strategies for what we would test in space.”
→ “ If we can operate on the edge with small language models like Gemma, we can reduce the amount of data teams send during downlinking by compressing it and federating satellite systems. That change can help us save satellite operators hundreds of thousands of dollars and also stack up other managed services from other software vendors to create a true marketplace in space.”
→ “With Google Cloud, Satlyt is creating a new way to scale compute beyond the limits of our planet.”
*Google Cloud products used:* Google Cloud, Google Kubernetes Engine, Gemini, Gemma, Gemini Flash 2.5, Google Workspace
Learn more:
→ Satlyt - Onboard AI & In-Orbit Computing https://satlyt.ai/
→ How Satlyt is building local AI for space with Gemma https://deepmind.google/models/gemma/gemmaverse/satlyt/
→ Make work less work with Gemini https://cloud.google.com/ai
→ Kubernetes, evolved: The foundation for platform builders https://cloud.google.com/kubernetes-engine
Google Cloud 다른 영상
New Way Now: ResiQuant cuts property underwriting processing times by up to 70% with Gemini
조회수 990
Gemini helps Stephen Curry's Underrated Golf Tour run smoother
조회수 6,451
Introducing Gemini for healthcare
조회수 44
AI doubled your output. Did it triple your incidents? | Dash0 on Google Cloud
조회수 1,040
How Gemini Agents Accelerate Drug Discovery 🧬
조회수 2,581
Orchestrate supply chains with Manhattan Transportation Planner in Gemini | Google Cloud
조회수 1,257
Connect SAP data and automate workflows with Gemini-powered AI Agents | Google Cloud
조회수 1,237
Accelerate sustainable finance with S&P Global’s Transition Finance Agent in Gemini
조회수 1,026