Dmytro Dzhulgakov
Dmytro Dzhulgakov is the co-founder and CTO of Fireworks AI, a generative AI infrastructure company focused on fast and efficient model serving. Before co-founding Fireworks AI in 2022, he spent nearly a decade at Meta AI, where he led the AI Frameworks team of more than 400 engineers and was a core contributor to PyTorch. He is also a co-creator of ONNX (Open Neural Network Exchange), the cross-platform deep learning model interchange format developed in partnership with Microsoft, Amazon, and others.
“Every serious AI application company will eventually need to train its own models to compete. Prompt engineering and off-the-shelf models hit a ceiling — the real leverage is baking your application's specific behavior, tools, and environment directly into model weights.”
Source→“Despite the full model being like one terabyte, not all the weights change every step... You can write a compression algorithm which basically leverages this property. And now you end up with maybe 20 times smaller delta than shipping the full model.”
Source→“Instead of crafting software directly, we went to crafting training data. Right now you're effectively crafting the evaluation rules, but that's still very important. You need to look at examples. You need to look at the data. You need to look at where your product fails and how to nudge the model in the right behavior.”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.