AboutPetra Vanlith
Where the writing comes from
Petra spent 9 years building production neural networks before writing a single tutorial. That gap between academic papers and working code is exactly what this blog addresses.
Most of her work focuses on the practical side - how gradient descent behaves on real datasets, why a 3-layer network underperforms a 2-layer one in specific conditions, what debugging a training loop actually looks like at 2am.
The gap between "I understand backpropagation" and "I can build something that trains" is where most learners get stuck - and that's the gap worth writing about.
What she actually covers
Neural Architecture
From feedforward networks to attention mechanisms - each article includes runnable code and explains exactly which hyperparameter to touch first.
Deep LearningTeam Tooling for AI Work
Distributed ML teams need tight communication. Petra covers how tools like Slack - including Slack for Windows and Slack for Mac - integrate into training pipelines and async review workflows.
- Slack notification hooks for model events
- Async code review via threads
- Channel structure for experiment tracking
Learning Resources
Every major article links to a curated reading list, dataset source, or reproducible notebook. No dead ends, no paywalled citations.
Open AccessA community, not an audience
team-chatting grew from a single Slack workspace where 14 people shared training logs and error traces. Over 6 years that group expanded into a full educational platform - the writing still reflects that original collaborative tone.
Readers ask questions that become articles. Problems shared in community threads shape the next deep dive.
If something in a tutorial doesn't work for your setup, the feedback loop is short. Petra reads every message sent to info@team-chatting.live and responds to technical questions within 48 hours on average.
Oksana Dreher
"The Slack integration guide saved our team 3 hours a week on manual status updates during model training runs."
publishing neural network content