At Guru, we believe the knowledge you need to do your job should find you. We integrate your company’s collective knowledge into a single source of truth and leverage AI to deliver relevant information to you in real-time within your work context. For customer-facing teams, that context is often a phone conversation. We built AI Suggest Voice to empower agents to focus on phone conversations and deliver amazing experiences to customers. AI Suggest Voice listens to customer-agent interactions and recommends verified knowledge to agents in real-time so they can confidently answer questions from customers, resolve complex issues faster and provide white-glove service that truly delights.
AI Suggest Voice is a cloud-native solution built on top of multiple Guru microservices working in tandem. In this talk, we will discuss the overall architecture of AI Suggest Voice and some of the lessons learned in making it all work in real-time and at scale. We will cover how we stream audio to our speech-to-text API over a persistent websocket connection and how we leverage state-of-the-art machine learning models and natural language processing techniques to detect speech, transcribe it to text in real-time, and use that text to suggest contextual knowledge.
The various components of AI Suggest Voice and its real-time at-scale requirements underscore both the challenging problem it solves and the transformational nature of the solution it provides for customer-facing teams.
Yev Meyer is a Senior Data Scientist at Guru, where he is helping the team fulfill the vision that the knowledge you need to do your job should find you. Using ML, Guru takes the “management” out of “knowledge management” and suggests verified knowledge to you in real-time wherever you work without you having to look for it.
Prior to Guru, Yev led and contributed to product data science teams at VC-backed startups Curalate and RJMetrics. He holds a PhD in Computational Neuroscience from Columbia University, where he studied dendritic processing and multisensory encoding in spiking neural circuits.