Store finder - performance optimisation with in-memory technique

Ashish Kumar (~ashish85)



Nowadays in the brand websites, it is common to find a store locator which helps you find a store nearby mostly powered by google maps. Using location service provider like google maps is all well and good unless you have a large (in order of millions) amount of locations to map and you need to do it frequently. Also using a location API service is out if we have inhouse custom location data. We have a new approach which saved us both time and money. We were able to find a nearby location from millions of locations in just a few microseconds.


Our test location database has around 1.5 million locations. For the purpose of this test, we performed the test on around 10,000 random locations and the results are as follows The time is for a single request. Benchmark Details

Join us for a chat, if you would like to know how we implemented it and how you can too in just under 60 lines of python code.


Basics of

  • Python
  • Numpy

Content URLs:

The repo is available in GitHub here

Speaker Info:

Data Engineer specialising in architecting and building data pipelines for massive data collection and training models for recommender systems. In the trenches no slideware only working software.

Id: 1570
Section: Core Python
Type: Poster
Target Audience: Intermediate
Last Updated: