Streaming Analytics
Resources
>
Case Study
>
Streaming Analytics
Business Problem
Leading operator of a two-sided marketplace that matches tractor owners to farmers, want to build state of art streaming analytics to identify tractor path and land ploughed.
Requirements
- Provide a cost-efficient solution for streaming data every minute by every tractor.
- Measure accurately the polygons of land that’s ploughed.
- High throughput data should be handled near real time.
Challenges
- Large volumes of data need to be processed by discarding noise.
- Identify transport path vs services provided.
- Non SQL database with some level of transaction management.
Solution
- Built a digital product that collects, cleans and analyzes streaming geo-location and vehicle data in real-time.
- The application algorithmically detects both the distance and area automatically and invoice the customer for tractor usage.
- The application is deployed and operated on AWS.
Business Outcomes
- 43% : Decrease in Discrepancies.
- 20% : Reduced manual corrections improved accuracy.
Tech Stack
- AWS Kinesis
- AWS EC2
- AWS RDS
- AWSLambdaDjango
- Geometry Engine
- GeoPandas
- PostGIS extensions
- Timescale DB
- MTurking
- Folium
- DynamoDB
- AWS Redshift