AI Quality Control (AIQC) Framework @ ISMB

[Layne Sadler]

As most of the folks interested in machine learning at the ISMB conference were on the MLCSB track, I though I'd share this talk about AI Quality Control (AIQC) from the BOSC track.

AIQC - accelerating research with an open source, declarative framework for deep learning
Why mining biobanks with pharma, I realized that the weaknesses of GWAS were the strengths of deep learning, but that AI was out of reach for most researchers - so I built AIQC to make it more accessible.

https://www.youtube.com/watch?v=suV5i-Y9tws

Documentation = https://docs.aiqc.io

• Declarative API that drastically reduces data wrangling
• UI for experiment tracking and inference
• SQLite based metastore

AIQC's pipelines support various types of data (tabular, sequence, image), analyses (classify, quantify, forecast, generate), and libraries (tf/keras, torch). You can think of it like the next level of abstraction of Keras in that it zooms out to orchestrate an end-to-end "raw data to insight" workflow.