hi everyone,
I ordered last month a Reachy Mini from @huggingface and while waiting for it to ship, an idea kept nagging me.
So I built https://t.co/ufhtt43K1v : a platform that helps people customize and upgrade their own robots, even if they are not developers or robotics experts.
Inference optimization isn't a one-time tuning pass. It's a loop: production logs → training data → better models → better production data.
Live webinar with Token Factory engineers on the full workflow 👇
Nebius is on stage at RAISE Summit in Paris this week. Hear from our leaders at four sessions across Jul 8–9, from why companies still underestimate the exponential to the token-per-watt era of AI infrastructure. Find the team at booth 16A for more info.
@skeptrune Get mintlify to work without bugs. Leave rl hot takes for rl researchers. It’s dumb to think oss models have seen all business tool use and workflows.
Today we’re launching Custom Speculator Training in Nebius Token Factory.
Train workload-specific draft models from your own data and deploy them alongside the base model.
Generic drafters are a good start, but production traffic is rarely generic 👇
Your CEO should be strong.
Your CTO should be wise.
Your COO should be wicked, cunning, of mysterious origins, fluent in the dark arts, blurry in pictures,