Expert data science: production, scale and responsibility

About this track

The expert track. A model in a notebook is a demonstration; a model people depend on is a system. Here you learn to frame problems that are worth solving, track experiments, serve predictions over an API, monitor for drift, test data and models, work with data too large for memory, handle text and time series, and take seriously the fairness and privacy duties that come with deploying a model that affects people. Finishes with the portfolio and interview work that turns all of this into a job.

Every lesson is written and hosted here on Yanjye — you never leave the site. Work through them in order, then sit the exam to earn your certificate.

Create a free account or log in to track your progress and earn the certificate.

Lessons

1. 1
   Framing: the work before the code

   Turning a request into a specification you can be held to.

   ~30 min
2. 2
   Reproducibility and experiment tracking

   Seeds, config, data versions, and knowing what produced a number.

   ~30 min
3. 3
   Serving a model: from pickle to API

   FastAPI, validation, batch versus real-time, and containers.

   ~30 min
4. 4
   Monitoring: models decay

   Data drift, concept drift, feedback loops, and what to alert on.

   ~30 min account needed
5. 5
   Testing data and models

   pytest for pipelines, data contracts, and behavioural tests.

   ~30 min account needed
6. 6
   Data too big for pandas

   Chunking, dtypes, Polars, DuckDB, Dask and pushdown.

   ~30 min account needed
7. 7
   Time series done properly

   Lag features, backtesting, seasonality, and the honest baseline.

   ~30 min account needed
8. 8
   Deep learning: when and how

   PyTorch fundamentals, and honestly assessing whether you need it.

   ~35 min account needed
9. 9
   Fairness, privacy and the duty of care

   Bias measurement, consent, minimisation, and the right to an explanation.

   ~30 min account needed
10. 10
    Portfolio, interviews and staying current

    What convinces an employer, and what to do after this track.

    ~30 min account needed