Are you ready at ToThePoint?

Machine Learning Constraint Solving

Build a constraint resolver in combination with a machine learning agent the solve an advanced employee scheduling problem.

Description of the assignment

Companies that schedule their work to be done in shifts, need to solve a common problem: how can we schedule our workers in multiple shifts adhering a couple of constraints and making sure all our shifts are filled in correctly.
Hard constraints, e.g. an employee can only do max 2 consecutive nightshifts, are easy to capture, but most of the time there are still multiple possibilities for the planning person to choose from.
The planning person has some soft constraints build trough experience in his head that he
will rely on to pick an ideal schedule among the suggested possibilities.

Goals

The idea is to capture these soft constraints in a machine learning model and see if it can be used to assist the planning person in making a better decision with minimal intervention.
We would build up proof that it is possible and can be put into production.
So basically, the project consists of the following tasks:
Implement a constraint solver for the hard constraints to solve a simple employee scheduling problem.
This results in a couple of possible solutions, which are then labelled using our predefined soft constraints to generate data.
A machine learning model will then be trained using that as input and as such learning the soft constraints.
Ultimate goal is then to combine the constraint solver and machine learning to provide the best solution among all possible solutions.

What you will gain

• You will learn to prototype and adjust minor errors
• You will learn real-time data captation and processing
• You will explore all kinds of possibilities of different sensors
• You will gain knowledge and experience in designing and producting a machine learning
solution
• That lovely feeling you’ll get knowing your product will in effect contribute to a neater trash
management system. It will effectively be used in our offices!

What you need

• Creativity and the will to succeed
• A motivated personality that can handle setbacks
• You see the bigger picture when it comes to making software and hardware work together
• You want to learn everything about real-time data and stream processing
• You can’t wait to learn a heck of a lot in a relatively limited time period.

Technologies you'll be using

Machine Learning
Python
Java
Deep Learning Framework (e.g. TensorFlow, Keras, ...)
Apache Kafka
Kubernetes
Docker
Flask - Rest API