Karan is the co-founder and CEO of Cartesia, an AI model lab building the next generation of real-time voice intelligence.

In this episode, Karan explains what all goes into building an AI Lab company; from massive datasets to pre-training to post-training to inference engines, relentless evaluation, and obsession with every layer of the stack.

We discuss why existing AI benchmarks are not very helpful, why latency is voice AI’s biggest enemy, and why maintaining context across long conversations is one of the hardest unsolved problems.

Apart from the cutting edge tech, this conversation explores Karan’s journey from researcher to founder, building Cartesia with a five-person founding team, raising capital, and realizing why first-time founders waste years focusing on things that ultimately don’t matter.

If you want a glimpse into the future of AI, watch this episode.

Watch all other episodes on The Neon Podcast – Neon

Or view it on our YouTube Channel at The Neon Show – YouTube