Job description

Background of the assignment

Cohelion provides data integration services to large multinational organizations. One of

those is a global air cargo handler that operates in over 300 airports worldwide. 5 years ago we had created a Machine Learning algorithm to create forecasts of operational tonnage volumes per station/customer/month. This was made using 8+ years of detailed historical data. A Python script was created by an experienced ML expert that has run since then.

The Challenge

Quite a lot has changed since then:

•more data is available

•the data quality has improved

•additional requirements have come in, for example to also take regional holidays into account

•Azure Machine Learning has matured

•and perhaps most important: the recent COVID pandemic has distorted many trends in historical data

The assignment/project

All of the above makes that a review/update is needed of the current algorithm that takes

the above into account.

We do see a possibility this assignment/project can be done within a project group of

2 students.

Expectations

We expect from this assignment:

•a new algorithm that performs better than the existing one (better means: forecast is

closer to actuals)

•ideally it would use Azure Machine Learning services

•sharing knowledge within our organization

Requirements

•experience with time-series forecasting (RNN or otherwise)

•experience with Azure ML stack (no requirement, but a bonus)

- Master's degree

- familiar with large, realistic (but anonymized) data sets

- Can create a new algorithm that performs better than the existing one (better means: forecast is closer to actuals)

Working conditions

•a very skilled and “gezellig” team to work with

•a beautiful view on the Maas river

•a healthy lunch (provided by us), regularly afternoon drinks and outings

•a fair internship compensation or hourly salary

- very professional and skilled colleagues