Appearance
Benchmarks
Are we faster than #AWS #AppRunner which is advertised for #MLOps? We needed to know - not guess. So we benchmarked with a detailed process-level profiling.
📊 Result: YES We are x 2.09 faster exactly - on 7 year old hardware.
We worked so hard on optimizing a 33 000 parameters cloud operating system, obsessively tuning performance to make this happen.
In the details we are x 2.35 faster on the ML tasks. So if you have ML Tasks there is no reason to use AWS anymore. Literally. That's less time, same resource usage, smaller footprint. Which means: lower cost.
And again: this is on aging hardware (7 year old). Imagine what happens when we match AWS on hardware while we overpass them on compute. (x 10 ?)
We don't have AWS's current scale, so we have to be better. Not a little better -- Much much faster! Stay in touch for more!