What Learners Found
When They Took the Time

Challenge

A senior data scientist at a logistics company had been asked to explore RL-based route optimisation but had no structured background in the field. Self-study had produced fragments that did not connect into a working model of the concepts.

Approach

She enrolled in the reinforcement learning module and used the hands-on environment — which resembles a simplified logistics graph — to anchor the concepts to her professional context from week two onwards. She brought specific questions about partial observability to every office hours session.

Outcome

She completed the module and went on to prototype a value-based approach for a subset of her company's routing problem. She described the course as giving her "a proper footing to be wrong from" — which we took as a reasonable measure of success.

Challenge

An analyst at a government-adjacent digital agency had been asked to review an AI procurement proposal and needed to evaluate the ethical claims being made by a vendor. She had no background in AI ethics frameworks and was aware that the vocabulary in the proposal was being used loosely.

Approach

She took the AI ethics and safety course over six weeks while managing the procurement review in parallel. The discussion sessions gave her the language to name what was missing from the vendor's documentation, and she used the case studies to develop a short internal framework for AI procurement review.

Outcome

Her framework was adopted informally across three other procurement processes at her agency. She completed the course with a clear view of where her knowledge had gaps and a reading list she continued to work through afterward.

Challenge

A backend developer at an e-commerce company had been moved into a data engineering role with limited handover. The pipeline he inherited had intermittent failures that the previous owner had managed through manual checks rather than systematic monitoring.

Approach

He enrolled in the data pipeline engineering course and brought his actual pipeline — anonymised — to office hours sessions from week three. Wei Liang helped him map the shared example structure onto the specifics of his situation rather than treating his questions as deviations from the syllabus.

Outcome

He redesigned the validation layer of his pipeline during the course and eliminated the class of failures that had been causing manual interventions. The monitoring he added has flagged three other potential issues in the months since. He described the course as "clarifying rather than adding".