Case Studies

Concurrent Scraping at Scale

These days, collecting information about anything and everything present in the public domain and using to find patterns that could prove crucial for various real world applications. In our use cases we had to iterate over not few, but aggregate information from tens of thousands of websites.

What I expected in this entire exercise was to reduce the painful running time of our existing processes. However, the output I got was robust in terms of User Experience, Monitoring, and a clean service layered architecture. This helped our organisation save a lot of money and time and was made one of the finest Automated pieces in our overall system.


High Performance Single Page Application with VueJs

Static HTML and simple web-pages are already a history now. The novel web applications are advanced and do a lots of functionalities. Also, the amount of data we show on a page and the relationship between them is growing large. For an application with a lot of related information scattered on a page, it is crucial to maintain the consistency between them. For a client, We had to develop one such app, which is extensively data-rich and we wanted it to be completely swift and seamless.

Francium Tech has really redefined the User Experience at a very fundamental level. Not only it is super fast, but is super intuitive as well.

Suresh Reddy, VP Engineering Services, Wilco Source

Customer Segmentation With Google Dataflow and TensorFlow

Identifying customer’s buying habit and categorizing them based on their shopping history is a key in retail domain today. It allows the organizations to do better promotion of products and improves sales. One of our clients wanted to do user-segmentation based on their purchase history over the past one year and based on the outcome they wanted to release promotional offers suiting each segment.

Francium tech were very thorough in their approach. Before turning over to them we had the same problem presented to different vendors, but none of their solutions were satisfactory. Not with these guys. Not only did they deliver well ahead of time, but exceeded our expectations! Using data analysed by this technique we were able to increase our promotions upto 4X.


High Volume background processing with Kubernetes and GKE

For one of our premium clients, we had to develop a highly scalable background processing engine based on Sidekiq. This is an intelligent system that parses documents and has an embedded machine learning component in it. The number and the size of the documents can vary drastically and the system should be efficient enough to scale for this change. In such a case, parallelism is the only solution for consistent delivery. So, we had decided to spawn multiple sidekiq jobs. But, the real challenge here lies in the confluence of these many modules and in managing them.

Before engaging with Francium Tech, we were very concerned about document processing speed as our existing one took almost 15 minutes to process and we could process only one at a time. However, they were able to lift both the barriers, the document transformation takes under a minute and there is no upper limit on how many could processed at the same time.