From data or ML engineering
Data engineers, analysts, data scientists and ML engineers who want to deploy AI systems inside customer companies.
What already counts
- Python and SQL at a professional level
- Pipelines, data quality checks and reconciling numbers
- Measuring models: test sets, precision and recall, thresholds
- Knowing how models behave and fail
What to add
- Shipping services: APIs, containers, retries and on-call habits
- LLM applications, tools and agents, which differ from training classic models
- Integration with enterprise systems, identity and security reviews
- Discovery with business users, and explaining trade-offs to executives
1Engineering for productionWeeks 1 to 3
Move from notebooks to services: call and expose APIs, handle failures, package and run what you build, and debug it from logs.
Show for it: A small API in a container that serves one of your past models or analyses, with retries on its own calls out and logs you can search.
2LLM systemsWeeks 4 to 6
Learn how LLM applications are built and measured: structured output, tools, retrieval and evals on documents.
Show for it: The claims extraction take-home with field-level accuracy, a confidence threshold chosen from data and a review queue.
3Customer systems and dataWeeks 7 to 9
Use your pipeline strength in the field: learn a customer's systems, move data safely between them and handle personal data lawfully.
Show for it: The HR sync take-home with a reconciliation report that a non-engineer can read and act on.
4Discovery and communicationWeeks 10 to 12
Practise the conversations: find the decision behind a data request, push back on a forecast nobody will use, and write for executives.
Show for it: Two Customer Room sessions where you turned a model request into a smaller first project, and a one-page update written for a business head.
5Switch and interviewWeeks 13 to 14
Present your data work as customer impact and prepare for the rounds where data people struggle: re-engineering code and customer role-plays.
Show for it: A portfolio page with two projects that show users, numbers and a running system, and a practised ten-minute walkthrough of each.