Our Projects

As part of the project we helped develop the influencer recommendation and relevance search algorithm and automated data pipeline for the social media analytics platform using Spark, Databricks and Elasticsearch.

Using individual level impression data and a machine learning model to predict affinity to banking products to manage the display ad budget on a individual lead level. The output was a automated data pipeline which ingested data from the ad system, scored leads using the machine learning model and uploaded audience data back to the ad system. Using Spark on Google Cloud Platform (Dataproc)

Online sales of motor insurance web leads - call centre prioritization predictive model

Fraud detection

Customer analytics and targeted micro campaigns (CLV, etc.)

Supply chain analytics and demand prediction

Building data science team and customers/ sales analytics, sales opportunities predictive model and distribution to sales agents

As part of the project KPIs for the areas of marketing, corporate sales and supply management have been defined and a set of interactive dashboards have been implemented for these KPIs. This enables Liftago to monitor the performance of their business and understand the key factors which contribute to the KPI development. This contains also an advanced prediction and planning model which the seasonality of the business and dynamics of the client portfolio.

MarketUP performs multichannel digital advertising campaigns for it's clients across a wide spectrum of ad systems like Google Adwords, Seznam Sklik, Facebook Ads, Adform, etc. The goal of the project was to ensure up-to-date online reporting on the campaign results using the Keboola and Microsoft PowerBI platforms across all systems both for MarketUP specialists and MarketUP clients. The goal was to increase the added value of MarketUP services towards it's clients and increase the efficiency of specialists by removing the need for manual reporting.

Operation risk management dashboarding and analytics

Python solution implementing our enhanced JLH-measure recommendation algorithm which is tailored to the specific situation of book recommendation where there is a huge number of books and sparse user rating data. Our algorithm combines publicly available external data (rating websites, etc.) with the internal data of the library.