The levels are described in more detail in a post at Towards Data Science here.

Entry level

Data handling

[ ] Small datasets
- Cues
[ ] Simple preprocessing
[ ] Image data
[ ] Audio data
[ ] Time-series data
[ ] Text data

Classic Machine Learning

[ ] Regression
[ ] Clustering
- Cues
[ ] SVMs

Networks

[ ] Dense Neural Networks
[ ] Convolutional Neural Networks
[ ] Recurrent Neural Networks

Theory

[ ] Mathematical notation
- Cues
[ ] Matrix operations
[ ] Regression
[ ] Clustering
[ ] Convolution
[ ] Simple metrics

General

[ ] ~1 million parameters
[ ] Learning the toolset
- Cues
[ ] Knowing the docs
[ ] Data analysis
[ ] Supervised data
[ ] Working with metadata files
[ ] Saving and loading models
[ ] Callbacks

Intermediate level

Data handling

[ ] Large datasets
- Cues
[ ] Imbalanced datasets
[ ] Complex datasets
[ ] Augmentations
[ ] Normalization
[ ] Generators
[ ] Data collection
[ ] Custom pipelines

Custom projects

[ ] Custom image project
[ ] Custom audio project
[ ] Custom time-series project
[ ] Custom text project

Networks

[ ] Large networks
[ ] Advanced layers:
- Cues
[ ] Custom layers
[ ] Language models:
- Cues
[ ] Generative networks
- Cues
[ ] Siamese Networks

Training

[ ] Transfer learning
[ ] Fine-tuning
[ ] Custom embeddings
[ ] Custom callbacks
[ ] Data-parallel training
[ ] Multi-GPU training
[ ] Custom training loops
[ ] Training in the cloud
- Cues
[ ] TPU training

Theory

[ ] Backpropagation
- Cues
[ ] Activation functions:
- Cues
[ ] Optimizers:
- Cues
[ ] Losses:
- Cues
[ ] Advanced metrics:
- Cues
[ ] Knowing common problems
- Cues
[ ] Advanced layers
[ ] Embeddings
[ ] Probabilities
[ ] Weight regularization:

General

[ ] ~100 million parameters
[ ] Unsupervised data
[ ] Creating splits
[ ] Code versioning
[ ] Experiment tracking
[ ] Hyperparameter search