About
My name’s Josh. I’ve spent the last decade building speech and AI systems. I got my PhD in automatic speech recognition from the University of Arizona, then worked at Mozilla on Common Voice and DeepSpeech. From there I co-founded Coqui, where we released open-source speech models that now have 6M+ monthly downloads. I’ve led applied AI teams shipping LLM and multimodal products, and I’ve published at NeurIPS, INTERSPEECH, and LREC (Google Scholar). Currently I’m a Founding Member of Technical Staff at Veris AI, building simulation environments for LLM agents. This blog is some of what I’ve learned along the way.
Even though I spend most of the day with my Speech / NLP / Machine Learning hat on, I’m still very much a linguist in that I love languages. I’ve spent a good amount of time studying French and Russian, and I’ve dabbled in a handful of others. Kyrgyz is a language I often use in my research, because it’s pretty cool!