SirionLabs HackFest

Problem Statement: Selecting a supplier from among thousands and doing a predictive performance analysis is a challenge which most of the client's face. Evaluating supplier performance based on multiple attributes would not only help clients mitigate risk but also help reduce costs. 

In order to achieve the same, we have provided some unstructured supplier performance data which may be millions of records, with multi-dimensional attributes from supplier rating, region, cost etc. 

Supplier performance can be defined as follows:

Formula using attributes. For example, Minimize [Cost] + Maximize[Rating] for upcoming year based on the historical data which can be executed within a specific duration [e.g., 30 days] for a specific service [e.g., Application Development] with minimum escalation 

As the expected outcome is meant for the future, all the numerical fields should be based on predictive computation as per historical data for each supplier.

Sample data set - 

There are a few things to keep in mind:

1. There are 500,000 supplier records
2. Each supplier will have mandatory following fields in every record:
• Supplier Name
• Country
• Period [Calendar year]
3. Other attributes are optional such as Cost, Rating. Most of the records will have these values. In case relevant values are not there, those records can be ignored.
4. Weightage can be given for each attribute (Sum of all attributes for a supplier should add to 100)
5. Each attribute will have only one value. So, there can be multiple records for each supplier for different services, regions, duration.
6. You can expect outcome to be dependent on maximum of 5 attributes

There is a multitasking scheduler system available, which can execute tasks/jobs in parallel at a given execution (1000) capacity. The system receives jobs, which need to be scheduled on the system. Each job is associated with Start DateTime, Finish DateTime, Value (1 to 100), Required Capacity (1 to 80), Frequency (in hours), DesiredStartDatetime and Execution Time(in Minutes)

Every job has to be started within 2 hours of DesiredStartDatetime. It is anticipated that jobs are already scheduled in the system, which may be again rescheduled to accommodate new jobs with higher value. Only those job instances can be rescheduled, which have not started.

Design a solution that will give maximum value for any given day. The solution should take care of job starvation and should provide basic reporting such as JobsForDay, ScheduledDaysForJob, and CapacityUtilizationForDay.

There are a few things to keep in mind:

1. All calendar days are relevant for scheduling jobs
2. The solution should generate alert messages once scheduler system capacity is utilized at 70%+ level.
3. There are 100 different types of jobs. You can configure Required Capacity (1 to 80) and Value (1 to 100)
4. It should pick a CSV file for importing jobs into the system for processing
5. If at any instant, execution capacity falls short, generate alerts in the next available time window.