Joe Fioti
6 things Joe uses or recommends, from 2 episodes.
Joe's top picks
What Joe uses or recommends most, ranked by how often and how strongly. Paid ads are left out.
Uses and recommends it · 1 episode
"We really like accelerators that have a big throughput focus. So, you know, our work on Trainium, for instance, we get excellent results out of Trainium. On our cloud we're doing something like 100 trillion tokens a month, and we're ramping a lot of that on Trainium. And we get pretty phenomenal, even against Blackwell Ultra, pretty phenomenal price performance on Trainium, and so we're really excited about that."
Uses and recommends it · 1 episode
"One of our first partnerships to make a compiler back end was with Positron... So we're big fans of Positron. We're actually ramping their Atlas chips in our cloud right now as well. So we love them."
All 6 of Joe's picks
- Trainium Uses and recommends it "We really like accelerators that have a big throughput focus. So, you know, our work on Trainium, for instance, we get excellent results out of Trainium. On our cloud we're doing something like 100 trillion tokens a month, and we're ramping a lot of that on Trainium. And we get pretty phenomenal, even against Blackwell Ultra, pretty phenomenal price performance on Trainium, and so we're really excited about that." Only Joe
- Positron Atlas Uses and recommends it "One of our first partnerships to make a compiler back end was with Positron... So we're big fans of Positron. We're actually ramping their Atlas chips in our cloud right now as well. So we love them." Only Joe
- PyTorch Uses it "We use a lot of tools that PyTorch provides where they're really nice about tagging the high level ops, and then we kind of track that metadata as it flows through our compiler. We use a lot of the standard PyTorch profiler to go and work out bottlenecks and issues." Only Joe
- SGLang Uses it "Then we live inside of the great inference engines of vLLM and SGLang. And so, the great solutions they already provide for speculative decoding, we're able to just go and directly piggyback on them." Only Joe
- egglog Uses it "Specifically, how we use the egglog engine, which powers our equality saturation." Only Joe
- vLLM Uses it "Then we live inside of the great inference engines of vLLM and SGLang. And so, the great solutions they already provide for speculative decoding, we're able to just go and directly piggyback on them." Only Joe