AI-assisted navigation for Kater ride-hail app
Kater was Vancouver’s first legal ride-hailing app. We worked with Kater to build their backend and solved various geolocation problems. We successfully designed and implemented geofencing and resource allocation algorithms that adhered to strict industry regulations. In this case study, we examine a geolocation problem that we solved using Elasticsearch and natural language processing.
Services
- AI Development
- Algorithms Design
- Cloud Architecture Consulting
- DevOps Automation
- Software Development
Technologies
- Amazon Web Services
- Docker
- Elasticsearch
- Machine Learning
- Natural Language Processing
Solution overview
Problem
When specific points of interest and addresses were entered as the destination, drivers were directed to inappropriate locations. Solving this problem required switching the mapping SDK, which was extremely expensive. Therefore, we needed to ensure that our solution provided competitive value compared to the cost of switching mapping providers.
Solution
A system that correctly positions the pins to ideal locations. For example, when selecting “Airport – Departure”, the routing software directed the driver to the correct location instead of the original coordinates that the routing SDK was returning.
Tailored maps
By leveraging various address search APIs to compensate for missing addresses and points of interest in the Lower Mainland, we created a tailored map experience for Kater’s users.
Points of interest
A dashboard allowed logistics personnel to create custom pickup and drop-off locations for different points of interest, which was particularly useful for the airport.
Natural language processing
A technical challenge was creating a solution that could deduplicate addresses from different data sources. We used natural language processing and custom-built proprietary algorithms to identify and deduplicate addresses with a benchmarked accuracy of over 99%.
Outcome
Our custom map solution solved a seemingly impossible problem. Leveraging open-source data and natural language processing, we created a comprehensive mapping solution highly optimized for ride-hailing in Vancouver.
