I continue my peer review of Machine Learning Zoomcamp midterm project with FORCE Lithology Prediction from Well Logs from spacelord99.
In general I find this a very well laid out project. It is clear that the developer has experience with geoscience and is uncovering meaningful data. As described below, I suggested that the problem statement could be better articulated
Problem description: A "problem" is not clearly stated. For example: "Interpreting lithology by hand from well logs is slow, somewhat subjective, and can vary between interpreters and across many wells", or something similar. So the intention of the project would state, "Reduce surveyor errors by supplying supplemental data", or something similar.
GitHub: spacelord99 FORCE Lithology Prediction from Well Logs
9. Cloud deployment
The only evidence is the Docker container running inside GitHub Codespaces, reached through a forwarded port. The README itself admits this is not permanent public hosting. A Codespace is a development environment, not a deployment target. The project has no deployment configuration at all: no fly.toml, Kubernetes manifests, or deployment commands for any cloud provider.
Excellent graphics and charts in the notebook itself. I can recommend showcasing the most interesting ones in the README, but this is obviously not critical.
5 Reproducibiliy
When I ran the command jupyter nbconvert --to notebook --execute midterm-project-notebook.ipynb --output /tmp/out.ipynb, I got the error "ModuleNotFoundError: No module named 'pandas'"
When I ran the command python train.py --output-dir artifacts , I got the error "ModuleNotFoundError: No module named 'joblib'".