As an MLE, our goal is to ship a scalable and safe model. Once the model becomes available to other teams on-demand, it’s important that our model can scale to meet demand, and that safeguards are in place to monitor and control for drift or degraded performance.
In this project, we'll be building an image classification model that can automatically detect which kind of vehicle delivery drivers have, in order to route them to the correct loading bay and orders. Assigning delivery professionals who have a bicycle to nearby orders and giving motorcyclists orders that are farther can help Scones Unlimited optimize their operations.
Note: This repository relates to AWS Machine Learning Engineer nanodegree provided by Udacity.
We’ll use AWS Sagemaker to build an image classification model that can tell bicycles apart from motorcycles. We'll deploy our model, use AWS Lambda functions to build supporting services, and AWS Step Functions to compose our model and services into an event-driven application.
We’ll use a sample dataset called CIFAR to simulate the challenges Scones Unlimited are facing in Image Classification. In order to start working with CIFAR we’ll need to:
- Extract the data from a hosting service
- Transform it into a usable shape and format
- Load it into a production system
We will use an AWS build-in image classification algorithm to train the model. Once the model is successfully trained, we will deploy it to an endpoint and configure Model Monitor to track the deployment. At the end, we will make an inference to test the model endpoint.
We'll write and deploy three Lambda functions:
- Save the code for each lambda function in a python script called lambda.py
- Take a screenshot of the working Step function
- Export the step function as a JSON file
