Kevin Lenzo

Timeline

1. 1995–1996

   The first Infobot

   Born in 1967, he writes the first Infobot in 1995, a Perl IRC bot that remembers URLs and runs in #macintosh on EFnet under the nickname 'url'. By 1996, he keeps a comprehensive list of Macintosh speech applications at Carnegie Mellon.

2. 1999

   The first YAPC and a White Camel

   He conceives and coordinates the first Yet Another Perl Conference at Carnegie Mellon in 1999, a grassroots meeting at $60 per person that draws nearly 300 people. He receives a 1999 White Camel Award along with Tom Christiansen and Adam Turoff.

3. 1999–2014

   FestVox and Building Synthetic Voices

   With Alan W. Black at Carnegie Mellon's Language Technologies Institute, he writes Building Synthetic Voices, a tutorial copyrighted from 1999 to 2014. The guide accompanies FestVox, their suite of tools for building new voices for the Festival speech synthesis system.

4. 2000

   Open-sourcing Sphinx and founding Cepstral

   In 2000, he releases the Sphinx 2 speech recognizer, originally developed by Xuedong Huang at Carnegie Mellon, as open source under a BSD-style license on SourceForge at LinuxWorld. In June 2000, he co-founds Cepstral, a Pittsburgh speech synthesis company, with Alan W. Black and other Carnegie Mellon scientists.

5. 2001

   Flite and the kal voice

   In 2001, he and Alan W. Black publish Flite at the 4th ISCA Workshop on Speech Synthesis, a small, fast run-time synthesis library for embedded systems and servers. He helps design the voice distributed with Flite and is its speaker; his voice also underlies FreeTTS and the cmu_us_kal_diphone Festival voice.

6. 2007

   Perl Foundation Director Emeritus

   In 2007, he leaves the voting board of The Perl Foundation, which he founded and served as founding President and then Chairman, and keeps the title of Director Emeritus.

7. 2008–2013

   An Apple text-to-speech patent

   He is one of four inventors, with Kim Silverman, Devang Naik and Caroline Henton, on an Apple patent application filed in 2008 for synthesizing speech from text strings tied to media assets, detecting each string's native language to produce human-sounding speech. The patent is granted in 2013.

8. As of October 2026

   Speech and AI at Duolingo

   As Speech Lab Lead at Duolingo in February 2023, he leads the team that keeps speech recognition and synthesis working across languages for speaking and listening exercises, and describes building custom neural voices for the app's characters with Microsoft. He works in Duolingo Research's AI and machine learning group.