Who We Are

Alex Sherstinsky and Eugene Mandel – we are decades-long collaborators who have worked closely together at six technology companies, with three focused specifically on conversational technologies. This deep partnership allows us to move quickly as a team and deliver results that stick.

We've witnessed the evolution of language processing technology firsthand. Alex's MIT PhD focused on machine learning well before it was called "AI." Together, we've built natural language and conversational AI products across multiple generations of these technologies – from "bag of words" and manually curated taxonomies to rule-based customer feedback systems for companies like Uber and Starbucks at Qualaroo, to human-in-the-loop systems answering customer support questions for major brands like Airbnb, Samsung, Microsoft, and LinkedIn at Directly, to full LLM-based conversation analytics engines.

Eugene built three conversational AI products over ten years. At Directly, he pioneered human-in-the-loop machine learning for customer support automation. At Loris, he led AI strategy for automated quality assurance and customer insights from conversation data. At RingCentral, he developed conversational intelligence products that analyzed millions of business calls, generating $11M in annual revenue. Throughout this work, he immersed himself deeply in the academic literature of Conversation Analysis.

Alex brings the research foundation and technical depth. His recent publications on LSTM networks and fine-tuning techniques demonstrate expertise in steering models toward specific domains – the key to applying LLMs effectively to particular business problems. At Predibase (acquired by Rubrik), he worked hands-on with fine-tuning large language models, combining theoretical knowledge with practical implementation experience. His significant contributions to Great Expectations, the leading open source data quality solution, address the foundational challenge of clean training data that reliable AI applications require.

We see two communities that haven't collaborated enough. Technologists have built impressive conversational AI algorithms, models, and products. In parallel, academic researchers have developed Conversation Analysis – a discipline at the intersection of linguistics and psychology that puts the study of human conversations on firm scientific ground, creating language to describe various aspects of conversations and their patterns.

These communities rarely collaborate. Traditional Conversation Analysis has been labor-intensive, relying on manual transcription and annotation that made it too slow and expensive for continuous business application. Meanwhile, conversational AI has lacked the scientific rigor to understand what actually drives successful business conversations.

LLMs represent a fundamental shift in capabilities for Conversation Analysis. For the first time, we can automatically transcribe, annotate, and analyze conversations at scale with human-like understanding of context and nuance. What previously required teams of trained analysts manually coding conversations can now be done systematically across thousands of interactions, making rigorous conversation science practically applicable to business problems.

However, the key to applying LLMs effectively to specific business problems is understanding the given problem space deeply and steering the models to align with goals and expert knowledge. This requires clean training data and cost-effective processes for human expert guidance – exactly the problems we addressed at Great Expectations and Directly with our human-in-the-loop systems.

Conversation Analysis is a scientific discipline that studies how conversations work, but until now it has been confined to academia because of its cost and complexity. Our platform brings this science to businesses by applying LLMs to analyze conversations with rigor and precision. It is not a generic tool that labels conversations as "positive" or "negative" sentiment. Rather, it provides a precise understanding of the conversational moves that drive business outcomes and clear methods to execute them effectively.

We're bridging the gap between rigorous conversation science and practical business applications, combining fifty years of academic research with modern AI capabilities to understand how conversations actually work – and how to make them work better.