The company Satlyt announced the engagement 8 million US dollars as part of the initial (seed) round of financing. The funds are aimed at deploying artificial intelligence and edge computing software directly on orbiting satellites to overcome the limitations of downlink communication channels and computing resources of spacecraft. The round was led by the venture fund Non Sibi Ventures (Houston) with the participation of TLcom Capital, Antler, Slauson & Co., Launch Africa Ventures та Enza Capital.
In the course of preliminary flight tests, the company successfully launched Google's open model DeepMind Gemma on the Momentus Vigoride orbiter, providing compression of error logs over than at 60 %. The software performs machine learning directly in orbit, processing the image, telemetry and syslogs locally, which significantly reduces data volumes, which need to be transmitted to Earth.
The current mission is carried out on the TakeMe2Space MOI-1A satellite, launched on a SpaceX Falcon rocket 9, and includes commercial and civil tasks, in particular research under the NASA STTR program, Stellerian's orbital surveillance software and TakeMe2Space application hosting. Over the next year, Satlyt plans to demonstrate distributed computing across multiple satellites with the goal of integrating on 20 % active orbital vehicles to 2030 year.
For reference: Satlyt was founded by former Google Cloud and SpaceX product manager Rama Afullo, has offices in Sunnyvale (California) and Nairobi (Kenya). Satlyt does not manufacture its own spacecraft, and develops a hardware-independent software platform, which turns disparate orbital computing equipment into coordinated virtual space data centers.
Source: https://satnews.com