Data Engineer 

Home Ex

A new kind of virus has been spreading around the world. Very little is known about it, yet the Israeli health care bodies have been able to collect some data from patients which they have made public in an attempt to get help from the country’s AI communities. 

Vanti has used this data to develop a new predictor, based on blood work done in hospitals, able to flag patients who are more likely to become symptomatic in the next 7 days.

What do we have?

 

  1. A production-level model, written in JS, that takes in testing results (numerical & categorical) and outputs a binary flag.

  2. The blood-testing machines' output, logged in their on-prem storage.

    1. Tests 1 : 250 are from Machine 0​

    2. Tests 251 : 500 are from Machine 1

    3. Tests 501 : 1000 are from Machine 2

    4. Tests 1001 : end are from Machine 3

  3. Each machine outputs its values asynchronously.​

  4. A dataset sample with Vanti’s predictions

 

 

 

 

Our goal

Provide a system architecture outline for 

  1. integrating our test into the machine’s digital test sequence

  2. Adding the test’s output as another test result per patient

  3. Presenting a dashboard to the hospital’s executives

  4. A CI\CD enabled workflow allowing a new model to be pushed into production every 1 week with 0 downtime.

  5. An ETL flow for asynchronous multi-stream data flow

 

Guidelines

 

Please submit 

  1. Architecture sketch

  2. Details on which frameworks used & why

  3. Cost estimation

  4. System capabilities and limitations analysis

 

Effort

  1. The entire exercise should take no more than 2h start to finish

 

 

Helpful things to consider before starting

 

  1. The answer to the exercise can be short as long as it’s well justified. We actually appreciate simple & elegant solutions 💪. 

  2. Make sure you are able to explain how your system works and to justify your choices.

  3. Does this problem resemble other architectures? If so, which ones and how do they resemble?

  4. If you’re entering hour 3, reconsider your approach 

  5. You can use any open source \ public knowledge you wish but, be sure you’re able to explain how it works and why you chose it.

  6. Stuck?         Call \ email us. Don’t be shy.

info@vanti-analytics.com

+1 (650) 843-9196 (toll free)

©2019 by Vanti.