AI, Speech & the Deepfake Era | Visar Berisha on Acoustics, Startups & Securing the Human Voice

From speech science to startup success.

In this episode of the Sound Cave Labs Podcast, we sit down with Dr. Visar Berisha, professor at Arizona State University, entrepreneur, and former co-founder of Aural Analytics. Visar shares how a speech processing project in college sparked his passion for acoustics and AI, leading him from MIT Lincoln Labs to defense industry roles, and ultimately to building a startup acquired by a digital health company.

We explore his work in speech as a biomarker for neurological conditions, how to commercialize research out of academia, and his recent focus on witness sensing to combat AI-generated voice deepfakes. Visar also opens up about balancing entrepreneurship with academia, mentorship, and advice for the next generation of problem-solvers in STEM.

Learn more about Visar’s research: https://visarberisha.github.io/
Follow Sound Cave Labs: https://soundcavelabs.com/

00:00:00 – Introduction & how a speech analysis project sparked Visar’s career
00:01:19 – PhD at ASU, MIT Lincoln Labs & defense industry work
00:01:59 – Becoming professor at ASU & role as Associate Dean for Research Commercialization
00:03:11 – Challenges of pushing faculty toward real-world impact
00:04:44 – Identifying problems vs. chasing solutions
00:05:22 – Use-inspired research & NSF I-Corps program
00:07:27 – Customer discovery & interviewing 500 people for real-world pain points
00:09:08 – Dual appointment at ASU in Engineering & Health Solutions
00:11:12 – Entrepreneurship journey: founding Aural Analytics
00:12:32 – Speech as a biomarker for Parkinson’s, ALS, Alzheimer’s & depression
00:15:32 – Finding product-market fit with pharma & clinical trials
00:18:53 – Growing Aural Analytics: SBIR, NIH grants & acquisition in 2020
00:20:41 – Lessons in pivoting, resilience & scaling startups from academia
00:26:18 – Speech analytics as a “neurological check engine light”
00:26:39 – Witness sensing paper: radar + speech signals to detect deepfakes
00:30:28 – Why radar + biology beats AI vs. AI detection
00:33:10 – Work-life balance vs. passion-driven careers
00:36:24 – Biological basis of speech, language variance & multi-language deployments
00:40:06 – Rapid growth of speech/AI research & international collaborations
00:41:45 – Voice banking for ALS patients & ethical AI use cases
00:43:13 – Advice to his younger self: enjoy the journey, not just outcomes
00:44:15 – Future of speech tech: silent speech interfaces & telepathic communication
00:45:21 – Reflections on MIT Lincoln Labs & closing thoughts