John Bohannon on translating machine learning research into products.
Subscribe: Apple • Android • Spotify • Stitcher • Google • AntennaPod • RSS.
John Bohannon is a Senior Director of Data Science and Head of Research at Primer AI, an end-to-end machine intelligence solution for textual data. We discussed their process of translating ML research into ML products, through the lens of the following examples:
- Zero shot entity recognition: zero shot entity recognition is a model capable of recognizing arbitrary entities without additional training.
- Inference triage and their open source project “baby bear”.
- Tools for detecting synthetic text.
- Text Summarization.
- End-to-end platforms for NLP applications.
If you are using or building NLP applications, make sure to attend the NLP Summit, a FREE virtual conference that takes place October 4-6.
Highlights in the video version:
- What is zero shot entity recognition?
Data sources, fine tuning, and labeling
Entity recognition and classifiers
What is inference triage and the story behind it?
Accuracy trade offs and evaluating summarization
ML and privacy
Synthetic text and synthetic text detection
End to end solutions vs. point solutions
Transitioning from a PhD to building products
Related content:
- A video version of this conversation is available on our YouTube channel.
- Foundation Models: A Primer for Investors and Builders
- Resurgence of Conversational AI
- Maarten Grootendorst: Unleashing the power of large language models
- Hilary Mason: Narrative AI
- Connor Leahy and Yoav Shoham: Large Language Models
- Omri Allouche: Machine Learning at Gong
[Image: Text Architecture by Ben Lorica; original photos from Unsplash, via Infogram.]