Welcome to my little corner of the internet. I'm Josh. I've been working on AI and language systems since 2011. A few things I've done: co-founded a voice cloning startup, released open-source models with 7.5M+ monthly downloads, lead applied AI teams, shipped LLM and multimodal SaaS and consumer products, shipped AI hardware, a PhD in automatic speech recognition, and lots of travel. Currently I'm a founding engineer at Veris AI, building simulation sandboxes for LLM agents.
Speech Recognition
-
Replacing Apple Dictation with Moonshine Flow (FOSS + local)
-
An Overview of Multi-Task Learning in Speech Recognition
-
My INTERSPEECH Schedule
-
Kaldi Troubleshooting Head-to-Toe
-
Kaldi Hyperparameter Cheatsheet
-
Kaldi nnet3 notes
-
Kaldi on AWS
-
Josh's Speaker ID Challenge
-
Seminal Papers in ASR
-
How to use an Existing DNN Recognizer for Decoding in Kaldi
-
How to Visualize a Word Lattice with Kaldi
-
How to Train a Deep Neural Net Acoustic Model with Kaldi
-
How to use an Existing GMM Recognizer for Decoding in Kaldi
-
Some Kaldi Notes
-
CMU-Sphinx Cheatsheet
-
Installing Kaldi
-
The CMU-Sphinx Speech Recognition Toolkit: First Steps
Speech Synthesis
-
Create New Voice with Ossian & Merlin
-
Getting started with the Merlin Speech Synthesis Toolkit
-
Let's make a Chuvash voice! : Moscow Higher School of Economics Speech Synthesis Workshop
-
eSpeak NG Notes
-
How to Add a Language to eSpeak NG
Machine Learning
-
How to Train practically any Model from practically any Data with TensorFlow
-
Maximum Likelihood Estimation of Gaussian Parameters
-
A TensorFlow Tutorial: Email Classification
Miscellaneous
-
Displaying Images in Claude Code
-
Watercolor Shaders for Ghostty
-
How we added Kyrgyz to Mozilla's Common Voice project
-
Some Linux Text Processing Notes
-
A List of Other Blogs
-
Some SoX(I) Notes
-
Soricut & Och (2015): Unsupervised Morphology Induction Using Word Embeddings
-
Installing Praat on Ubuntu - getting sound to work
subscribe via RSS