At PolyAI, we’ve always recognized that effortless, fluent conversations between customers and AI agents over the phone requires a platform that has a strong foundation in speech recognition. As a proud member of the NVIDIA Inception program for startups, we saw the opportunity to build an industry-leading speech recognition model, PolyAI Owl, fine-tuned specifically for more accurate understanding and better responsiveness in real-world, complex customer conversations.
We chose to build our speech recognition models on NVIDIA Riva to access state-of-the-art foundation models which have already been trained on vast amounts of speech data. This enables our teams to build an ensemble of speech recognition capabilities to adapt for different languages or accents, noisy environments and other factors that affect understanding of conversations over the phone.
We’re excited to further enhance the scalability and flexibility of our speech recognition stack with NVIDIA NIM microservices. The PolyAI platform handles millions of conversations each month, and the ability to react to peaks in demand and call concurrency on a moment’s notice for our enterprise customers, without compromising on the quality of conversations, is critical for us. With NVIDIA NIM, we’re now able to more efficiently provision the right speech recognition capabilities, with the confidence that we can respond rapidly to any spikes in demand across our diverse base of customers, use cases, accents and languages.
“PolyAI is committed to helping enterprises show up as the best versions of themselves through our incredibly lifelike and helpful conversational AI agents,” said Nikola Mrkšić, co-founder and CEO of PolyAI.
“With NVIDIA NIM and Riva, we’re strengthening the foundation of our speech recognition stack, ensuring that our enterprise customers get unmatched accuracy, speed, and flexibility for conversational AI at scale.”
Learn more about PolyAI Owl or our Agent Studio platform.