The Bombardier challenge has gotten our attention due to its implication with data, customer services, optimization and as it turns out machine learning. It also captured our interest because of the business & industry sociability potential.
What it does
OptiMaint is a web platform that allows rail companies (and possibly others) to manage their trains (fleets) in such a way that optimizes the maintenance and availability schedule.
How we built it
The backend is built using Python (with Keras & SQL). The program can take a continuous flow of data or manual input such as diagrams. The frontend is built using react.
Challenges we ran into
Aggregating enough human data for the machine learning algorithm to find the best overall routes for each train. Keeping UI & UX as intuitive as possible for people without a technical background to use it. Because no machine algorithm is perfect and we cannot allow for mistakes to happen as these may cost the client money, resources, time and their image.
Accomplishments that we're proud of
Developing a machine learning system with a constraint mechanism on top at a hackathon. Adapting years of human generated data to be used by a machine in order to determine the optimal train assignments with less resources than a human might have.
What we learned
We understood the importance of communicating with our clients in order to understand and adapt our product to their needs and specifications (eg maintaining an AGILE workflow). We also learned the importance of planning a good business model that will help us have a better overview of our product which we can then improve for our clients.
What's next for OptiMaint
First step would be refining the product and getting it from a MVP stage to a industry grade product/service.
OptiMaint as modular and available as possible so that more companies could integrate it into their systems.
A more refined and detailed micro-servicing maintenance approach and additional feature.