RCA programme for a major Brazilian logistics company
RPA, a Python orchestration daemon, ML models and a web interface. Client-estimated annual gain of R$ 500 million.
What it was
A cognitive automation (RCA) programme for one of the largest logistics companies in Brazil, covering stock usage, forecasting of incoming orders and load assembly.
What I did
I led technical delivery from a six-month proof of concept through to production, spanning multiple project phases. The solution combined:
- Automation Anywhere RPA
- a custom Python orchestration daemon
- machine learning models supporting operational decisions
- a web interface for client integration
The bin packing problem
The most enjoyable part was a variation of the bin packing problem. As a mathematics enthusiast, I reached a workable solution two ways:
- a custom algorithm that iterated through the stock, discarding unfeasible solutions
- a genetic algorithm, to optimise the ordering and which cart each pallet went on
That work became the paper Usage of a genetic algorithm for optimizing stock usage, presented at GECCO 2020.
Outcome
Client-estimated annual gain of R$ 500 million, with improved inventory flow, faster delivery to large and mid-sized retailers, and reduced time for load assembly and validation.