Renen Hallak

Lin Qiao of Fireworks AI said speed and price are going to be crucial differentiators for AI apps, and touted the approach of using open-source LLMs that are customized and closely tied to the application for post-training and inference.

“The continuous loop is the secret sauce,” she said.

Some companies have come up with excellent products but can’t offer them because the compute costs are too high, she said. “Product-market fit and having a viable business are two separate things” in the current moment, she noted, and everyone needs faster performance and lower cost.

Renen Hallak of VAST Data said model-builders and enterprise customers alike need tools to be able to store increasingly massive amounts of both structured and unstructured data, and easily interact with it. GPUs are getting “much hungrier for parallel access,” he said, and the data infrastructure will need to keep up.