I started in biotechnology and scientific research: lab work, bioinformatics, biologics manufacturing, and production. Those jobs were about evidence, controls, and being able to repeat a result.
That carried over. A pipeline has to run the same way each day. A forecast has to survive backtesting. A dashboard is only useful if the metric definition stays stable.
The objects changed from assays and batch records to tables, models, and reports. The constraints changed too: cloud cost, latency, and how Finance or Risk defines revenue. Validation stayed the main habit.
If you are making a similar move, I would talk about the problems you already know how to bound. What counts as evidence, what would prove a claim wrong, and what has to keep working when you are not watching it.