Kumparanian

Kumparanian is a set of workflows that optimize Kumparan's data scientist hiring process. It cuts down 1-2 working day(s) submission review process to just less than an hour.

If you are our candidate, you need to install kumparanian using following command (we highly recommend to install inside virtual env, like venv):

python -m venv <your_env_name>

source <your_env_name>/bin/activate

pip install kumparanian 

consult its help command:

% kumparanian ds --help
Usage: kumparanian ds [OPTIONS] COMMAND [ARGS]...

  For Data Scientist role.

  Before you submit your trained model, you can verify your trained model
  using the following command:

  $ kumparanian ds verify YOURMODEL.pickle YOURFILE.pickle

  YOURMODEL.pickle should contain your trained model, and YOURFILE.pickle
  should contain the necessary preprocessing components such as the vectorizer 
  and label encoder.

  Use the following command to evaluate your trained model against your test
  dataset:

  $ kumparanian ds evaluate YOURMODEL.pickle YOURFILE.pickle test_file.csv

Options:
  --help  Show this message and exit.

Commands:
  evaluate  Evaluate the model
  verify    Verify the model

  If you found any issues, feel free report it at:
  https://github.com/kumparan/kumparanian/issues

then read our assessment and you should be good.

Subsequent sections are not required for candidate as it intended only for project's documentation purpose.

Kumparan's Model Interface

The first component of Kumparanian is a Kumparan's Model Interface. We've designed an interface for Machine Learning model that allows us to design a problem to have deterministic result.

The model interface contains 3 required methods: train, predict and save. The candidate will solve the assessment test by implement the train and predict methods. We provide save method to helps the candidate to save the trained model.

Read more about the Kumparan's Model Interface.

Kumparanian CLI

The second component of Kumparanian is a kumparanian. This CLI will help the candidate to verify and test their model while also help our team to evaluate the candidate's trained model.

kumparanian can be installed via the following command:

pip install kumparanian

To get started, run the following command:

kumparanian --help

If you found any issue, please open new issue here kumparanian/issues.