I’ve noticed that "Machine Learning Engineer" roles are now 70% software engineering and 30% modeling. Should my first project include a full MLOps pipeline with GitHub Actions, DVC for data versioning, and Prometheus for monitoring? Or is it okay to just focus on the Jupyter Notebook and the model results for now to show I understand the algorithms?
3 answers
To be honest, a Jupyter Notebook alone won't get you hired in 2026. Every hiring manager has seen a thousand notebooks. If you take that same model and put it into a CI/CD pipeline where it automatically retrains when new data arrives, you’ve instantly moved into the top 5% of candidates. Use DVC to show you can version your data just like your code, and use a tool like MLflow to track your experiments. This proves you can work in a professional team environment where "reproducibility" is more important than a one-time high accuracy score in a notebook.
Is there a risk of over-complicating a simple project with too many MLOps tools, potentially distracting the interviewer from the actual ML logic you used?
Notebooks are for research; scripts are for production. If your portfolio is all .ipynb files and no .py files, you're signaling that you aren't ready for a production role.
Exactly, Gregory. Alicia, definitely prioritize a clean, modular code structure. It makes your project look like real software rather than a school assignment.
Ryan, it's a balance. You don't need a 20-tool stack for a simple classifier. Just pick one: versioning (DVC) or deployment (Docker). Showing you understand that a model needs to live in a "system" is what matters. You want to show you are an engineer, not just a researcher who doesn't know how to deploy code.