Last XKE we had a brainstorm about this. Here are my notes:

New DE assessment needs to be about:

• problem solving
• distributed systems
• working with other peoples code

Timing:
Current timing suggestion (8-16h) is too long and too much to ask for a person to fit into his/her weekschedule. Need to be brought down to 8-10h

Implicits:
We discussed that the readme for the assessment has some implicit expectations in it (like we expect to see unit tests etc)
One thing needs to be made more explicit though:

• Handing in the code 24hrs in advance. Stretch that it doesn't need to be the final version

A reminder 24hrs in advance would be helpfull. Also not plan the presentation on a Monday to prevent stealing valuable coding time from the candidates

Data:
The fact that we have 2 datasets with different formats is a good thing. Keep that

Streaming:
If a streaming use case is added it needs to have the possibility of instant and verifiable results/visualizations

Multiple people who now the old xebia photoshop assessment liked the fact that you will be given a piece of code and had to/could do multiple things:

• Make it run
• fix some bugs
• Add some feature(s)

We discussed the use of the divolte shop project we have as a demo for divolte for this purpose:

Divolte shop:
Has:

• Docker containers
• Elasticsearch
• Kafka
• Redis
• Grafana
• Java code
• Python code
• A web front end
• Multi arm bandit

It hasn't:

• Spark

It isn't broken

We can create specific branches with specific broken parts, problems, bugs that fit the skillset of the candidate.

We need:
A way for the candidate to work on the code without seeing ther peoples solutions.
Investigate the use of private repo's/branches, instruct or github classroom

Actionpoints:

• Investigate repo/platform to use
• Add spark stuff to divolte shop
  • Streaming divolte events from kafka
  • Have 2 large divolte event datasets with mutliple formats or different avro schema's
• Create different type of bugs in the code to address different skill sets
• Investigate possible features to add
• Is docker-compose the right setup or do we need a cloudbased environment available