REFERENCES

THE CHALLENGE

The digital bank wanted to have an holistic view of the main processes including digital enrollment and the life-cycle of the main products, as mortgages, loans, credit cards, etc. The information is disperse and operational KPIs are not defined.

THE SOLUTION

Keepler defined with the bank +300 operational KPIs and designed the data flows from both internal and external sources in order to calculate and display the KPIs. Keepler also designed a comprehensive set of dashboards with self-analytics capabilities.

THE RESULT

The bank used this information to enhance operational processes, also fixed the calculation of indicators that was previously misinformed or wrongly calculated.

THE CHALLENGE

The exploration environment in an on-premise Cloudera installation has to be upgraded to accommodate more data and more users.

THE SOLUTION

Migration of a 250Tb Data Lake to the cloud, including data ingestion, processing using Spark and data consumption using a tailored data scientist environment. The new data exploration platform is built using native public cloud services and fully automated. It provides services to +200 data scientists.

THE RESULT

Undisclosed savings in Cloudera licenses.

THE CHALLENGE

With a team of 200 data scientist and an on-prem Cloudera platform, the customer was struggling  to create new data environments and keep costs in line.

THE SOLUTION

Keepler designed a Big Data platform to mirror the Cloudera capabilities with native services as S3  and EMR and migrated 250Tb of compressed data to cloud. Also automated the ML pipeline with DataLab environments that can be launched on-demand. 

THE RESULT

Reduced the time of delivery of Data Scientist environments from days to minutes. Improved the performance of current Spark processes.

THE CHALLENGE

The insurance company receives more than 100k emails per month related to different processes and requests. A team of experts must read all emails and forward them to the appropriate team. This process is costly and prone to human mistakes.

THE SOLUTION

Keepler designed and developed an automatic email classifier that is able to OCR attachments (Textract), extract topics, intents and forward the mail automatically to the appropriate team (Sagemaker).

THE RESULT

+90% precision in the classification.

Reduction of 20% of the mail analyzed due to the detection of spam.

THE CHALLENGE

With several brands and e-commerce platforms, our client wanted to obtain an holistic and unified view of customers and sales in all brands. Using this information, the client wanted to increase online sales by targeting every customer with a personalized recommendation.

THE SOLUTION

Keepler deployed a Big Data platform uploading customer interaction data (Google Analytics) and sales identifying customer using cookies, loyalty card, etc, and associating the customer with sales from the ERP backoffice platform. Even brick&mortar sales data were used to further personalize the recommendations.

THE RESULT

Internal users have a unique view of the ecommerce platforms in all brands and sales were boosted thanks to the combination of recommendation in the portal and a tailored newsletter automatically sent to known customers.

THE CHALLENGE

The steering knuckle is a key and expensive component of the car. If there isn't stock of this piece the whole factory has to stop. The car manufacturer wanted to forecast when the knuckles will arrive at the factories in two Spanish cities.

THE SOLUTION

Keepler integrated an IoT solution based on Cellnex and devices attached to the racks of knuckles. Each device can track the position of the rack and if it is moving.

THE RESULT

As the application is rolling out, trucks are being monitored and stock can be better adjusted. By providing a steady flow of steering knuckles the client can keep an average of 2.000 cars produced per day per factory.

THE CHALLENGE

Increase average price of oil products (gasoline, diesel, etc) without affecting the demand.

THE SOLUTION

Design training and deployment of +1.200 data models to dynamically adjust gasoline price in each gas station according to demand elasticity, weather, city events, and a business rule engine.

THE RESULT

Increase of 2 cents of Euro of the average gas price.

THE CHALLENGE

A Bedbank is a wholesaler of hotel rooms. Travel agencies browse room and hotel availability through the Bedbank application. The customer wanted to reduce the volume of searches by recommending a personalized set of hotels to each customer.

THE SOLUTION

Keepler deployed a data product that create white-lists of hotels and locations for every travel agency according to a segmentation based on a customer profile and historic bookings and searches. The segmentation and the white-lists are periodically and automatically calculated in order to further reduce the look-to-book ratio.

THE RESULT

Look-to-book ratio were reduced in a 10% thus reducing the operation costs of the bedbank accordingly.

THE CHALLENGE

The customer wanted to enhance customer experience when checking in and reduce the need of personnel in reception desk in peak season.

THE SOLUTION

Keepler designed a check-in process based on a mobile application that capture the faces and documents of the guests before the trip, and through a tablet device in the hotel, detects the identity through face recognition and automates check in.

THE RESULT

This initiative is part of a pilot.

THE CHALLENGE

Workplaces as refineries are places where workers must wear specific security material. It is hard to detect when workers are not wearing this material.

THE SOLUTION

Design and deployment of a model to detect the use of protective equipment (helmet) and deployment in a edge computing device (camera with GPU processing device).

THE RESULT

This initiative is part of a pilot.

THE CHALLENGE

LNG cryogenic pumps are key components of the LNG transportation and any performance deviation has a huge impact in the service.

THE SOLUTION

Design and productivization of a deep learning model to detect and predict performance deviation in LNG cryogenic pumps. This model is automatically retrained and is available via API.

THE RESULT

Ongoing project.

THE CHALLENGE

The customer is willing to identify water use patterns (specifically fraudulent) and also automate the detection and fixing of gaps in measures.

THE SOLUTION

Keepler designed a Big Data platform in the cloud that can process and store smart-meter data. A set of data models were also deployed to manage gaps and identify anomalies.

THE RESULT

The data quality of the entire process was improved, increasing the accuracy of the customer's invoice estimation. During the project, the identification of potential fraudulent cases, leaks or flow-reversal cases helped the customer's operations area to detect issues in advance.

THE CHALLENGE

The client is willing to increase service stickiness by providing value added solutions based on IoT devices connected to the mobile and internet router.

THE SOLUTION

Keepler designed and deployed a IoT real time analytics end event-driven platform that enables the intelligent correlation of events to launch actions (alarms, other IoT events, etc) and obtain insights from the customers.

THE RESULT

The client has provided its customers with an advanced app that allows them to monitor the state of their home and its habitants, with services ranging from intrusion detection to assistance to the elderly.

THE CHALLENGE

Although new data on COVID-19 is available daily, in a pandemic situation it is relevant to know the evolution of the spread to make preventive decisions.

THE SOLUTION

Keepler developed and implemented an interactive dashboard to visualize and understand in a better way the key drivers of the pandemic and its implications. Based on machine learning technologies, it allows developing models at a regional level to evaluate and forecast the course of the pandemic.

THE RESULT

Forecasts are strongly influenced by the reliability of the data. Having an early warning system capable of obtaining a holistic view of the evolution of the pandemic and the incidence of the virus among Spanish regions, allows to detect changes in the distribution of these data and make decisions based on them.

THE CHALLENGE

Maintenance of pump equipment is one of the most important tasks associated with the operation of process plants and it was implemented either on a routine basis or after the failure of equipment. Enagás wanted to change this procedure and anticipate possible failures by observing abnormal behaviours in the data.

THE SOLUTION

Through the proposed machine learning model, the process evolves to a data-driven predictive maintenance system. In addition, an API has been developed so that the trainings, inferences or updates of the models can be performed easily by business users.

THE RESULT

This solution provides Enagás a model for every pump helping their technical team to evaluate the functioning of every pump individually,, reducing costs and intervening only when strictly necessary. Thanks to the API, users can perform multiple trainings/inferences faster, making this task more agile.

THE CHALLENGE

Meet the EU GDPR requirements in terms of data privacy, applying pseudonymization techniques to PII data stored in a ~300TB AWS S3 Data Lake.

THE SOLUTION

Built an advanced, cloud-native, and event-driven architecture based on AWS EMR, AWS Lambda, AWS SNS, and AWS S3 to de-identify historical data and on-the-fly data making use of direct integrations with operational services (orchestration flow and data catalog). The platform also allows users to request de-identify or identify operations on-demand and it also provides a complete layer of monitoring and alerts.

THE RESULT

Data users will not be allowed to see or misuse PII information that could potentially lead to an unexpected data breach, while they can keep working as usual with the encrypted data.