Neocrym is a collection
of open source software
and industry research.
Neocrym was originally founded by James Mishra as an indie software company applying artificial intelligence to the music industry.
We built a lot of tools to help us train a massive number of machine learning models on audio data.
Now we are giving away out tools for free.
Here is a list of our major open-source software projects and our research.
Babycat is a library that makes it easy to decode and resample large batches of audio files in parallel.
It is written in Rust and has bindings for C, Python, and WebAssembly.
- Source Code: github.com/babycat-io/babycat
Provose is the easiest way to deploy to Amazon Web Services.
It makes it easy to deploy hundreds of individual AWS resources--containers, databases, TLS certificates, DNS rules, etc.--with just a few lines of code.
When you use Provose to deploy two resources that need to talk to each other--like a container and a database--Provose automatically configures the relevant security groups and IAM policies to ensure that those resources can securely communicate.
ScalarStop is an experiment tracker for reproducible machine learning research.
It makes it easy to log machine learning training metrics to a local SQLite or PostgreSQL database, making it easy to track the results of a distributed hyperparameter search.
ScalarStop also makes it possible to hash and cache trained machine learning models--saving compute resources by ensuring that the same model is not accidentally trained twice.
Click Track is a newsletter about technology and the entertainment business.