This proposal outlines the scope of building an ML system that can assign volunteers to disaster relief using an efficient algorithm.

This project will take 9 weeks to implement and cost $96,000 in fees.

Building the ML system involves three parts -

• Data exploration and preparation

• Modeling

• Model evaluation and validation

The team will be staffed with a mix of data scientists and developers from Pirates’s network of startups. One senior data scientist, one data scientist and one developer will work on the project and other resources will be pulled in as necessary to complete the project on time.

Given a set of volunteers and a current disaster, how do we achieve the optimal assignment of volunteers to the current disaster?

First, we need to break down this problem statement into two parts -

1. Given a set of volunteers and a current disaster

2. Optimal assignment of volunteers to the current disaster

Feature Engineering

Feature Engineering involves a bunch of activities performed on the features of the dataset making them more appropriate for their use in modeling.

Data Modeling

We model the data to generate key data objects which can be represented as vectors. Then we analyze the interaction of these vectors.

Model Fitting, Evaluation

This stage involves understanding how the Association, Volunteer and Disaster vectors exist in their own dimensional spaces and how these vectors interact.

The work product of the project will be to create a rudimentary web app which will use the ML system to enable ARC to give a disaster (with its attributes) as input and get a list of volunteers ranked by match scores.

We will also provide detailed documentation showing the results of evaluating the data models against test data and visualizations the data models’ performance.

ARC and Pirates will work together to create a set of success metrics to measure the ML system

We don’t use PIIs as much as possible and if there is an absolute need to use PIIs they will be anonymized in the datastore using hashing algorithms.

QIs will be protected by bundling groups of QIs in addition to anonymization.

In addition we use techniques like Randomized Data Ordering and Differential Privacy for further protection.

Pirates will undertake any additional tasks necessary to complete the project successfully even if it takes longer than the 9 week timeline.

If the project execution takes longer than 12 weeks, Pirates shall forego the final installment of the payment and yet deliver the work product successfully.

The team for this Pirates will comprise of data scientists and developers from Pirates’ network of startups. The initial team will comprise of a senior data scientist, a data scientist and a full stack developer. As the project progresses, additional resources will be added by Pirates as deemed necessary to successfully execute the project.