AI and ML Technologies

on Amazon Web Services

Our vision envisage a foreseeable future in which companies base their operations and decision making on data. It is what we call “Data Products” that rely on the cloud and AI techniques as cornerstone.

Data Science

Sagemaker makes data science operations of building, training and hosting of Machine Learning models much more easy. With the addition of functionalities for the operation of the models, such as deployment, a very complete work environment is achieved, with more integrations and functionalities.

Data processing and management

In order to ensure optimal ML models, it is very common that data preprocessing is required to carry out ETL process. This can be supported by services such as EMR, Batch or Glue, with the benefit of the latter, which also allows data to be cataloged.

Analytics

Multiple services on AWS allow data analysis, both in the exploratory phase, and in the rest. Some examples are: Athena, to query data in S3; Kinesis Analytics, for streaming data analysis and QuickSight, for data visualization.

Image and video AI services

AWS Rekognition allows the recognition of various types of objects in images and video, especially facilitating facial recognition. Textract is used to extract data and text from scanned documents (OCR).

Voice and text AI services

Polly lets you transform text to speech and Transcribe does the reverse. Translate does translation between languages, Comprehend is used for NLP (natural language processing), and lastly we have Lex for chatbots.

Other AI / ML services

AWS has other high-level AI services such as: Personalize, to make recommendations; Forecast, for prediction or Fraud Detector, to detect possible online fraudulent activities.

#MachineLearning Cases of Success

CASE OF SUCCESS: CEPSA

PERSONAL PROTECTIVE EQUIPMENT
DETECTION USING VIDEO ANALYTIC

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

INDUSTRIAL
CROSS
IoT
Cloud

CASE OF SUCCESS: CAF

MULTI-CLIENT BIG DATA PLATFORM FOR
TRAIN MAINTENANCE OPTIMIZATION

The client is willing to increase the maintenance services ROI by leveraging the data generated by the trains. Trains sensor data is massive (2Gb per unit per day) and complex to analyze.

transformación digital en CAF
TRANSPORT
Big Data
ML
Cloud

CASE OF SUCCESS: ENAGÁS

Anomaly detection in cryogenic pumps

By observing abnormal behaviours in the data, maintenance can be scheduled before there has been a failure which would lead to loss of production and spoilage of gas.

ENERGY
ML
Data Science
Cloud

CASE OF SUCCESS: CAM

Modeling the spread of Covid-19 infection

Although new data on COVID-19 are available daily, information about the biological and epidemiological characteristics of COVID-19 remain limited, and uncertainty remains around nearly all parameter values.

PUBLIC SERVICES
ML

AI is the best technique to extract value from data

At Keepler we work for companies to generate value from data through Data Products. These products take advantage of the flexibility of the public cloud and machine learning techniques to address all phases of a data project with an MLOps approach based on automation and monitoring in each step towards a complete ML system: build, integration, testing, deployment and productive management.

This allows us to build solutions that are more scalable, faster, smarter, and adapted to the needs of each company.

At Keepler we understand technology as a tool with which to create value, data as its basis and AI as the best technique with which to extract it. Whether it is a proven solution or new approaches to identify, we will accompany you provinding successful technical solutions thanks to our experience and know-how.

Iterative and incremental deliveries of value

Taking into account the nature of the project, we propose an approach based on CRISP-DM (Cross-Industry Standard Process for Data Mining). This methodology comprises an open standard process model endorsed by AWS in its guide to good architecture practices (Well-Architected Framework) applied to Machine Learning (ML Lens). As a complement to this framework, for proper project management, we incorporate risk management events, planning, review and continuous improvement through a framework such as Scrum. The conjunction of both frameworks allows us to make iterative and incremental deliveries of value throughout a Machine Learning project in continuous collaboration with the client.

AI use cases

Customer Management

Customer management is becoming more complex: more demanding customers, more points of contact and more data. AI is the best way to gain a true understanding to improve interactions with our customers, choosing the right channel and the right content. AI helps organizations improve the user experience with information coming from their systems or other channels.

Recommendation systems

The number of products and services grows day by day. At the same time, interactions through digital channels are increasing (e-commerce, digital platforms such as Netflix, Spotify). It is important to have solutions that allow you to offer the most appropriate product or service at the right time, combined with business rules managed by business users.

Time series and demand forecasting

The AI is able to predict upcoming sales, website visits, electricity or fuel demand in a specific area. The forecast in time series obtains the information about the origins of the forecast, which are the events with the most impacts and the “what-if” scenarios.

Natural Language Understanding (NLU)

The algorithms of the NLU allow us to manage a huge volume of data generated in different documents. Some examples of these capabilities are: Detection of highlights in texts, classification of intention and taxonomy, sentiment analysis.

Voice, text and image recognition

Advances in AI bring new possibilities. Audio transcription and image and video recognition have many applications such as damage detection or recognition of individuals. The large number of algorithms and techniques provided by different frameworks and services designed by cloud providers, simplify and accelerate this process.

Optimization in model deployment

One-click production deployment and model training is possible, including: Automated A/B testing, variable optimization, automated training, model accuracy monitoring.

We talk about AWS

Entendiendo tus propios modelos predictivos | T3chFest 2019
Detecting “things” with edge computing and the cloud | T3chFest 2019
Keepler | Webinar | Los datos que explican el covid-19 en España
Aplicando Agile en productos de datos para mejorar el éxito del proyecto
Simulando la vuelta a la normalidad: la teoría de grados y el Covid-19
Predicción de series temporales con redes de neuronas 
Let’s talk! 

If you want to make the move to the public AWS cloud, contact us and we’ll talk.