Willow Voice outsources AI training complexity to CoreWeave's specialized infrastructure
The dictation app maker is leaning on CoreWeave's reinforcement learning platform to handle the operational burden of model training, freeing its team to focus on the core technology.
Building AI products often forces startups to become infrastructure operators at the same time. Willow Care Inc., which develops the Willow Voice dictation application, selected CoreWeave Inc. as a partner not merely for raw computing power but for hands-on operational support. The product merges speech recognition capabilities with language processing technology.
Lawrence Liu, co-founder and chief technology officer at Willow, explained the appeal of outsourcing this layer of complexity. "The great thing about CoreWeave's [reinforcement learning] infrastructure is we don't have to worry about anything under the hood," Liu said. "The only thing we really need to focus on at this point is eval alignment."
Liu shared these insights during an appearance at the Fully Connected event, speaking with theCUBE Research analysts Dave Vellante and John Furrier. The conversation touched on CoreWeave's reinforcement learning platform, the role of compact language models in post-processing speech, and Willow's ambition to reduce keyboard dependency.
Training a specialized voice interface
Willow has invested heavily in fine-tuning its own speech recognition model using real-world audio samples, including whispered speech in noisy environments like coffee shops. Converting recognized speech into usable text requires more than just accurate transcription, Liu noted.
"I think the main part of our secret sauce is we have this really, really small LLM that post-processes everything that you say," Liu said. "Now it's able to format everything correctly, take your inputs and change it to exactly what you would've written."
Currently, Willow's computational demands center on training, though inference workloads are accelerating as user adoption grows. This trajectory aligns with CoreWeave's strategic emphasis on inference capabilities.
https://www.youtube.com/embed/fsshYkbI83c?start=1&feature=oembed
"Within two years, we want it such that you don't even need keyboards anymore," Liu said. "You can just speak and your computer understands."