Telia - Area of Data Services
Business Challenge:
Undefined foundations for storing and handling all the incoming data from on-prem to the cloud, resulting in lack of confidence from application teams to move to the cloud. And hence impacting innovation, cloud transformation and competitiveness
Project Description:
Move services allocated on-premises to the AWS cloud, for cost reduction and scalability. Architect and design a robust data lake system which can be easily extended with new functionality, governed (GDPR) and maintained.
How We helped:
We helped in developing and maturing the data lake platform in the AWS cloud, responsible for storing and running complex batch processing jobs (ELTs) as well as streaming jobs for all the company data.
tech stack
VALUE
Moved partially away from expensive on-prem servers and expensive licenses. Telia reduced OPEX, reduced significant CAPEX and prepared a solid and versatile platform for the cloud and ML functionalities that speed up the development lifecycle by 30%.
Project 02
Telia - Area of ML Services
Business Challenge:
In their quest to stay competitive and innovative, they needed a ML platform so that data scientists could experiment and release functionalities. Data scientists and business analysts came from different teams in different countries with different technical backgrounds meaning that a multi purpose and intuitive platform was needed in place.
Project Description:
Develop and mature the ML platforms so application teams can use them confidently for their data analysis processes. They should be able to experiment, release complex models to stage and production.
How we helped:
We worked with data scientists from different teams on releasing ML platform functionalities so application teams could leverage the ML platform for doing their data analysis processes. Our principal contribution was to seamlessly integrate the data foundation platform with the ML platforms by architecting and defining the usage of Sagemaker Feature Store. We also contributed actively in MLOps, extending, supporting and maintaining the aspects of the ML platforms to make it intuitive.
tech stack
VALUE
More than 20 data scientists from different countries and business units could use the platform for experimenting, developing and releasing predictive models for detecting when a client (out of the 25M) will churn based on existing behavioral data or internal processes of incidences and how they impact. Telia could act upon it proactively and secure for more months the client by offering discounts, gifts or vouchers. Imagine reducing the churn rate from the average industry in Telco from 12% to 10% by proactively improving your services.
Factmata
Business Challenge:
They had started developing their SaaS with an ad-hoc data lake system. It was difficult to extend, scale and integrate with new API providers. They were a data-intensive application leveraging databases that were not optimized at all for data consumption by the SaaS, taking seconds for simple requests, impacting user experience.
Project Description:
Re-design and implement a robust data system that can ingest, transform and store massive amounts of data from different sources like Reddit, Twitter and Instagram. The system should support batch processing to efficiently compute and load data into curated layer for ML processing and into the databases the processed data for consumption by the SaaS backend.
How we helped:
We helped them build a solid data system that was scalable by designing and creating a server-less ingestion system that could scale on-demand. Our core contribution was around data consumption and ML batch pipeline processes.
tech stack
VALUE
We optimized by 60% the infrastructure costs and decreased by 80% P95 metric for API requests to the underlying relational databases, making a drastic positive impact in the user experience when using the SaaS, hence reducing churn rate from 17% to 11% and keeping existing users.
Ready to unlock new value for your business?
