Part 4/12:

To address latency issues, Akash emphasized the benefits of edge computing—processing data directly on satellites to trim down the volume of data needing transmission. For example, instead of downlinking entire 900MB images, onboard systems can identify and transmit only relevant insights like flood extents or road networks.

However, implementing AI at the edge in space is non-trivial. Satellites face numerous constraints: limited power, thermal management challenges, storage restrictions, and the inability to physically upgrade hardware post-launch. Akash shared that early efforts, like embedding NVIDIA GPUs in satellites (dating back to around 2019), faced hurdles due to high power consumption and heat dissipation demands.