Machine Learning Engineer Assesment
Every month, there are thousands of events active on TicketSwap websites. These events range from music festivals to concerts to sporting events, stand up comedy shows and more. To help our users find the events they would like to attend, we try to extract the event lineup, as well as any additional information, from the event titles.
Part 1
Inside the data directory there are two files:
- A file containing a list of event titles, event_titles.txt
- A list of artists from our database, artists.txt. For the sake of this assignment, we do not differentiate between artists, stand up comedians, football clubs, or festival names. They all are listed together in the same artists.txt file.
We would like you to build a solution, given an event title, it would return the potential lineup (artist names) or equivalent (football teams, festival names, etc).
Your solution should consider the following:
- Not all artists are listed in our database. Think of ways to extract ones not in our database yet.
- The same artists can have different spellings, think Pink vs P!NK.
- The same name can refer to multiple things, think Arsenal (football club) vs Arsenal (band)
Bonus questions
- What additional information can you extract from the event titles? What additional business value do you think this information can provide?
- Can you differentiate between your confidence levels for when different artists are extracted?
Part 2
Once your solution is ready, other parts of our product will need to communicate with it. What is the best way to make your predictions available for the rest of the product to consume? Will your solution differ in case the predictions are needed in real-time vs offline batch jobs? What are the different deployment options you can think of?
Please build a very simple prototype for making your predictions available on demand to our different microservices.
Part 3
Your prototype proved to be successful and you want to push it forward. How would you communicate with the following teams and what evidences will you try to provide to convince them about your solution:
- Content team. This is the team that currently adds line-up to events manually, and this project will help them automate this part of their job
- Marketing team. This is the team that needs to have line-ups for events to be able to promote the events with popular artists.
How would you approach other teams to put your solution in production, and make sure they can consume your predictions successfully? For example, the front-end and back-end team who will need to integrate with your solution? Can you think of other technical teams that you will need to align with?
Finally, will you need to collect additional data for your future iterations on your solution?
Project Structure
MACHINE-LEARNING-ASSESMENT
│ README.md
│ requirements.txt
│ .gitignore
|
└───data
│ │ artists.txt
│ │ event_titles.txt
│
└───src
│
└───test
In this repo you'll find three directories and three files. Inside the data directory, you'll find two datasets. Any other dataset that you consider necesary for the project, should be added in that folder.
In the src directory you should put all your service & model related code. Use the test directory to add all the files you need containing test for your code. Please, make sure to add all the dependencies needed to run your code in the requirements.txt file.
Handing in the assessment
Remember, it's important that you clone this repo before starting and do as many commits as needed, as clear as possible. You will not have push permissions to this repo, so you will have to do it in your own git repository (either local or using your own github account). When you're finished developing the application and did all the git commits, you can zip the entire folder (don't forget the .git), and send it to us in reply of the mail you got with this assessment. If there are issues sending the zip file, please use a service like https://wetransfer.com to upload the zip file there and send us a link to download the files.
Please, do not share the results of this assignment with anyone.
Best of luck to you!