Implementing AWS Services for Training, Enrichment, and Inference
Question
You work as a machine learning specialist for a farming corporation that wants to use in-ground soil sensors together with enrichment from geolocation, rainfall, and other weather information for the growing area to help identify crop growth stages.
They want to use the crop growth information to increase yield and produce more product year over year.
They also hope to increase the crop quality through this effort. The machine learning models that you build for this solution will analyze various growing conditions, such as temperature and humidity.
So the farming corporation can schedule watering appropriately for the area. What collection of AWS services would you use to implement a solution that first trains your model, then gathers the information from the in-ground sensors, then enriches the sensor data, and finally deploys the model to run inference on connected devices in the field?
Answers
Explanations
Click on the arrows to vote for the correct answer
A. B. C. D.Answer: A.
Option A is correct.
SageMaker is used to create your model and train it initially.
IoT Core sends the sensor data to IoT Analytics for enrichment and analysis.
The pre-trained model is deployed into the field using IoT Greengrass so you can perform ML inference using the enriched data on the farm local devices in the field.
Option B is incorrect.
You could use Kinesis Data Analytics to analyze your IoT device data streams.
Still, IoT Analytics is built specifically for analyzing highly unstructured IoT data.
So, it is a better choice.
Option C is incorrect.You could use Kinesis Data Streams to stream your IoT device data, but you would have to write lambda functions to perform the enrichment step.
IoT Analytics is built specifically for analyzing and enriching highly unstructured IoT data, so it is a better choice.
Option D is incorrect.
Inference Pipeline is used to define and deploy pretrained SageMaker algorithms.
Inference Pipeline does not have the IoT inference integration that IoY Greengrass has.
So, IoT Greengrass is a better choice for this problem.
Reference:
Please see the AWS IoT Greengrass ML Inference overview, the AWS IoT Analytics overview, the Amazon Kinesis Data Analytics overview, and the Amazon SageMaker developer guide titled Deploy an Inference Pipeline.
The correct answer to this question is option B, which involves using SageMaker, IoT Core, Kinesis Data Analytics, and IoT Greengrass.
Here is a detailed explanation of why this option is the best fit for the given scenario:
SageMaker: Amazon SageMaker is a fully managed service that provides developers and data scientists with the ability to build, train, and deploy machine learning models at scale. It is a natural choice for training the machine learning models required to analyze the various growing conditions and schedule watering appropriately. SageMaker provides pre-built algorithms and frameworks, which can speed up the development of machine learning models. Additionally, SageMaker allows easy deployment of the models on the cloud or on edge devices.
IoT Core: AWS IoT Core is a managed cloud service that enables devices to connect securely and reliably with cloud applications and other devices. IoT Core can be used to gather information from the in-ground soil sensors and other IoT devices in the field. It also enables secure and efficient communication between devices.
Kinesis Data Analytics: Amazon Kinesis Data Analytics is a fully managed service that can analyze streaming data, such as the sensor data collected from IoT Core. Kinesis Data Analytics can enrich the sensor data with additional information, such as geolocation and weather information, as required by the farming corporation. Kinesis Data Analytics can also be used to preprocess the sensor data and make it ready for machine learning models.
IoT Greengrass: AWS IoT Greengrass is a software that extends AWS functionality to edge devices, allowing them to act locally on the data they generate while still using the cloud for management, analytics, and storage. In this solution, IoT Greengrass can be used to deploy the machine learning models built using SageMaker to connected devices in the field. This enables the devices to run inference on the sensor data locally, without relying on the cloud for every decision.
Option A, which includes IoT Analytics instead of Kinesis Data Analytics, is not the best fit because IoT Analytics is a service that is designed to analyze IoT data stored in the cloud, rather than analyzing streaming data.
Option C, which includes Kinesis Data Streams instead of Kinesis Data Analytics, is not the best fit because Kinesis Data Streams is a service that provides a platform for streaming data processing but does not include the analytical capabilities required for enriching data or preprocessing it for machine learning models.
Option D, which includes Inference Pipeline instead of IoT Greengrass, is not the best fit because Inference Pipeline is a feature within SageMaker that enables building and running pipelines for inference on batch and streaming data, but it is not designed for edge devices like IoT Greengrass.
