TekBay, an ISO/IEC Certified IT Resourcing & Cloud Consulting Company and an AWS Consulting Partner, helped Prixa Technologies build an agro-based predictive analysis tool with AWS Sagemaker for their client.
About the Customer
Prixa Tech is an IT services Company based in Nepal that primarily builds Advanced AI/ML, Analytics, CMS, Advertising, and Custom Software Solutions. They emphasize innovative, efficient, and optimized innovations to cutting-edge pioneer solutions adapted to the client’s needs.
Business Challenge
The requirement was to develop an Agro-based system that attempts to forecast crop demand & yield to optimize supply decisions. To build the system, Prixa primarily faced challenges to create an automated architecture and machine learning models on AWS to:
- Analyze and predict the yield based on collective data from previous seasons, fertilizers distributed in that area, soil type & weather forecasts.
- Suggest the best crops that could attain maximum yields in specific areas based on the yield potential and demand & supply of the crops in the market.
Why AWS?
Prixa chose to build the solution on AWS as it provided the needed machine learning services & solutions for the forecasting tool. The ML models were constructed on Amazon SageMaker, as it allows for a fully managed way that helps create, design, train, tune, and deploy machine-learning models quickly.
Why Did They Choose TekBay?
TekBay has a capable team of certified engineers with expertise in architecting and building MLOps solutions on AWS. Prixa chose to engage with TekBay as we proposed a solution that met their requirements. As an AWS Partner, we could provide AWS services & support locally in Nepal with a high level of competency in their domain.
Solution Delivered
Based on the client’s requirements, we proposed an architecture for an automated MLOps pipeline. The system uses a serverless architecture, which has benefits in terms of deployment edge and affordability, and is fully automatic and event-driven.
Data is observed, captured from multiple sources, and stored in Amazon RDS and S3, then used for model training with SageMaker. To build this system, we employed the following AWS services:
- AWS Step Functions for orchestrating the various jobs within the pipeline and incorporating logic for model validation.
- Amazon S3 for initial data storage/data lakes, storing flat-file data extracts, source code, model objects, inference output, and metadata storage (the initial feature store might also be a relational database, e.g., Amazon RDS, and alternatively, Amazon DynamoDB used to store metadata for some use-cases).
- Amazon SageMaker for model training, hyperparameter tuning, and model inference (batch or real-time endpoint).
- AWS Glue or ECS/Fargate for extracting, validating, and preparing data for training jobs.
- AWS Lambda for executing functions and acting as a trigger for retraining models.
- Amazon CloudWatch for monitoring SageMaker tuning and training jobs.
Architecture
Outcome
This automated solution forecasts the crop type for seasonal cultivation for maximum yield. For Model performance monitoring, this solution uses CloudWatch and step functions to retrieve evaluation metrics within the pipeline.
The following forecasts are generated by using this AWS solution:
- Total crop yields during a season
- Yield possibility of crops
- Maximum & minimum-yielding crops
- Best Crops for specific areas & seasons
