Using AI for Speech-Language Therapy

Using AI for Speech-Language Therapy
Lead Researcher: Venu Govindaraju
Institution: University at Buffalo

Millions of children face speech and language challenges that can affect their ability to communicate, learn, and succeed in school. While early intervention can significantly improve outcomes, many communities struggle with shortages of speech-language pathologists and other specialists.

Using Empire AI’s advanced computing resources, the team is building multimodal large language models (mLLM) capable of delivering real-time, interactive speech-language therapy. The technology is being developed for use by the National AI Institute for Exceptional Education, a $20 million federally funded initiative led by UB focused on creating AI systems that help identify and support children with speech and language-processing challenges.

The Challenge

Speech and language disorders can have lasting effects on academic achievement, social development, and self-confidence. Although early intervention is critical, many schools and families face limited access to specialized services due to workforce shortages and rising demand.

Effective therapy also requires individualized attention. Clinicians must continuously assess a child’s responses, adapt activities, and adjust instruction to meet changing needs. Delivering this level of personalized support at scale remains one of the biggest challenges in special education.

The Approach

The University at Buffalo team is developing a suite of AI-powered tools that  provide real-time speech-language therapy. Unlike traditional AI systems that simply answer questions, these models are designed to act as an adaptive partner capable of planning, delivering, and modifying interventions based on a child’s behavior and responses.

The multimodal systems combine language understanding, visual awareness, and contextual reasoning to engage children in dynamic learning activities. As therapy sessions unfold, the AI will monitor interactions, interpret responses, and adjust its approach to support each child’s learning goals.

How Empire AI Makes this Possible

Creating a real-time AI system that can understand language, interpret visual information, reason about interactions, and adapt its behavior requires substantial computing power.

Empire AI provides the advanced infrastructure needed to train and evaluate these sophisticated multimodal models. Researchers can leverage large-scale computing resources to process complex datasets, refine model performance, and test the system across a wide range of educational scenarios.

This access enables faster development and experimentation, helping move promising AI technologies from research environments into real-world educational settings more efficiently.

 Potential Impact

For educators and therapists, these tools will serve as valuable supplements that extend the reach of limited professional resources. For families, they could create new opportunities for children to receive individualized support both in school and at home.

Most importantly, the project demonstrates how advanced AI can be applied to improve lives. By helping children develop critical communication skills and receive support earlier, this research has the potential to strengthen educational outcomes, expand opportunity, and improve quality of life for children across New York State and beyond.