Giters

ruodingt / mle-challenge

Whispir - E2E ML Engineering Challenge

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

🤖 E2E ML Engineering Challenge

Setup and Submission Instructions

  • you'll be using the following tools to complete the challenge

    • Python 3.8 (& pip)
    • Docker
    • Any other tool/libary you decide to use

  • Setup:

    • fork this repository and clone the fork onto your local machine
    • create a new branch for your solution
    • feel free to commit & push (as you usually would) to this newly created branch as you work on your solution

  • Submission: When you are ready to submit your solution, you can either

    • open a Pull Request into the main branch of the original repository (from your branch on your fork). We will get a notification once you open your PR and we will review your solution. Feel free to also let us know that you've submitted it if you'd like. OR
    • zip your repo and email it to us if you prefer to keep your solution private. PS: please zip your .git as well as we'd like to see how you've approached the solution. (zip -r my-solution.zip . should do the trick)

  • And please don't hesitate to reach out to us if any of these instructions are not clear or if you have any questions.

Good luck!

Challenge

Part 1: Training

  • In this part of the challenge, you'll code, train & operationalise an ML model to predict the Wind Power Generation for a given day.

  • Dataset: Wind Turbine Data. This dataset contains historical data for a wind turbine. It contains various weather, turbine and rotor features. Data has been recorded from January 2018 till March 2020. Readings have been recorded at a 10-minute interval.

  • Task:

    • Build and Train a Machine Learning model to predict Active Power (and we can expect other columns to be provided as input features).
      • Feel free to use any ML library you like (e.g. sklearn, tensorflow, pytorch, etc...) or any off the shelf models.
      • We won't be judging your modelling/Data Science skills. We are more interested in your general problem solving, engineering, and applied ML skills.
    • Operationalise training
      • We except your code to be production grade & deployable, so we look at things like repo structure, code quality, tests, debuggability, reproducibility etc...
      • Please automate & containerise parts (or all) of your building, testing & training processes as you see fit.

Part 2: Serving

  • In this part of the challenge, you'd be serving your model as an API.

  • Tasks:

    • Build a REST API to serve your model
      • Pick your best model from the previous part of the challenge
      • Wrap a REST API around your model (using tools like flask, django, fastapi, etc...). We'd prefer if you don't resort to off the shelf serving tools like ternsorflow serving or torch serve as we want to see that you can build a simple model serving system from scratch.
      • Feel free to use any ML library you like (e.g. sklearn, tensorflow, pytorch, etc...) or any off the shelf models.
    • Operationalise serving
      • We except your code to be production grade & deployable, so we look at things like repo structure, code quality, tests, debuggability, reproducibility etc...
      • Please automate & containerise parts (or all) of your building, testing & serving processes as you see fit.

Hints / Tips

  • We have intentionally left out parts of the tasks open ended and ambiguous to not lead you to a particular approach. However, if you'd like hints, tips or suggestions to get started, we are happy to share them with you. They are currently in the form of a markdown file called hints.md in a password protected zip hints.zip.
  • Do reachout to us if you'd like access these.

References

About

Whispir - E2E ML Engineering Challenge