Granulation process optimiser

The brief

Each pharmaceutical trial costs material and time, so finding good process settings in 10 trials rather than 100 saves real money. Build a simulated granulation process, then use Bayesian optimisation to choose the next trial and compare it with grid search and random search. Good looks like: the optimiser reaching the target in far fewer trials, a record of every trial, and the limits of the simulated process stated.

What you will build

1. Build a simulated process with noise
2. Define quality targets and constraints
3. Run grid and random search as baselines
4. Implement Bayesian optimisation
5. Compare trials needed to reach the target
6. Write recommendations for a real process

Free, open data to build it with