This article was shared in our internal BIML thread and prompted a reflection from my days at Shazam:

Before VoiceAI and LLMs… there were birdsongs. This is a great story, success on many levels, of one of the world’s first, far-reaching Sound Recognition apps (and of Machine Learning, and, yup, AI).  

Every month while sitting in London building Shazam, I’d see a few requests for “can you do Shazam for Birds?” I loved this question. 

But when I pushed our Chief Scientist, Avery Wang, “Please Avery, why can’t we do it?!” the reasoning, which I listened to closely, would grow a core operator skill: knowing when to push for innovation, when to back-off, and when to double-down on your core value prop. We were music discovery and needed to create an entirely new user behavior.

Branching into a huge science project would have been derailing at a stage when the then start-up was fighting to stay alive. 

Thankful there were incredible scientists, researchers and builders who did it with full-focus. Cornell Lab of Ornithology is the OG.  And their Merlin App is extraordinary. I’ve been a fanatical user for years and love seeing how the database is helping conservation and biodiversity efforts. 

And a crazy full circle: my colleague at BIML (Berryville Institute of Machine Learning), Harold Figueroa was a core creator of the technology behind Merlin. ✨