Unlock cutting-edge AI/ML research for enterprise voice tech
We investigate and innovate solutions to complex challenges in speaker recognition, liveness detection, and AI–empowering enterprise contact centers to deliver smarter, more secure customer experiences.
Bridging the language gap between humans, AI, and enterprises
We're advancing AI research in linguistics and systems to develop scalable solutions that excel in real-world complexity. Our advanced models help businesses enhance CX, improve automation, and bolster defenses against identity-based attacks, paving the way for more secure, efficient interactions.
Better spoof detection for synthetic voices
Voices can be cloned with just 30 seconds of audio, creating challenges for systems and agents to distinguish real from fake. Our advanced model quickly adapts to new voices and changing conditions–even with limited training data–providing businesses with more reliable protection against AI-driven voice fraud.
Our focus areas
Bespoke automatic speech recognition
Custom ASR models designed for your specific domain or application.
Conversational AI and intelligent agents
Human-like virtual agents that understand and resolve customer needs.
Enterprise use of large language models
Surface insights, automate workflows, and drive better decision-making.
Be the first to know about our latest innovations
Join us at upcoming events where we share our latest research findings and innovative solutions on speaker recognition, liveness detection, and LLM integration in enterprise systems.
From research to real world
Enhancing transcription accuracy
ASR systems often struggle to accurately transcribe audio, especially in noisy environments. By leveraging LLMs, we select the optimal ASR output, reducing word error rates and improving transcription quality.
Going beyond the surface of sound
Using sparse autoencoders (SAEs), we uncover hidden features in audio data to enhance speech recognition and voice authentication systems, enabling businesses to break down complex data into insights.
Boosting voice authentication in multi-speaker settings
Traditional systems have difficulty differentiating speakers in group settings. Our new technique increases speed and accuracy of multi-speaker detection–enabling better voice biometrics, diarization, and forensics.
