Edge Computing
edge computing, drone et intelligence artificielle en bout de chaîne

Edge Computing

Fraud detection, cyber defence, purchase prediction, drones, driverless vehicles, video games, personal assistants, human resources, predictive maintenance and artificial intelligence have all become part of our daily lives, are growing exponentially and revolutionising industry linked to systems architecture and value creation.

At the same time, given servers’ inability to cope with the explosion of data as the use of sensors becomes widespread, edge computing allows connected objects to perform certain tasks currently carried out by data centres and provides a decentralised infrastructure by giving the objects themselves some intelligence. By moving computing power to the edge of Internet of Things networks, network infrastructure traffic is lessened, latency reduced and energy management improved.

Edge computing is the result of the progress made in miniaturising sensors and developing new integrated circuits embedding artificial intelligence systems directly into chips. Thanks to artificial intelligence techniques, decision-making can happen almost in real time without the need to go back and forth to the cloud.


With its expertise in the various technological building blocks that create artificial intelligence, Astek is at the heart of innovation in the field

Automatic language and speech processing

Used for speech recognition, translation and automatic text generation, automatic support, chatbots, sentiment analysis, etc.

Our expertise:

  • Transcribing human speech into a format that computer applications can use
  • Analysing text to facilitate understanding of the structure and meaning of sentences, feelings and intention
  • Using unstructured data

Automatic learning

Used in anti-spam filters, optical character recognition, search engines, computer vision, etc.

Our expertise:

  • Extracting knowledge from raw data
  • Designing effective supervised/unsupervised learning algorithms according to how each model family is used
  • Automatic massive data processing and predictive analytics
  • Implementing neural networks (machine learning, deep learning)

Image processing

Used in scene reconstruction, video/motion tracking, item recognition, image restoration, etc.

Our expertise:

  • Extracting and structuring information for indexing images
  • Image search using keywords and content
  • Man-Machine interaction
  • Recognising shapes in image sequences

Artificial intelligence at the heart of Astek innovation

Technological developments and the increasing flow of data enable businesses to gain additional knowledge and anticipate the needs of their clients more accurately.

In partnership with an insurance company, Astek developed a way to optimise financial management of the cost of natural disasters via real-time assessment of the severity of climate events and their scope using geo-localised information from social networks.

Square Predict, a software factory called Scub Foundation for Hadoop, uses sentiment analysis on Twitter and will be able to automatically collect and interpret tweets to determine the proportion of positive and negative opinions on a given subject and display these metrics in real-time.

As well as its superior information processing capability, the system will become more intelligent as it reads tweets, using an automatic learning mechanism (machine learning/Lambda architecture). Thanks to lexical and semantic analysis, it adds new words it comes across to its dictionary and extrapolates a positive or negative feeling associated with the word. One of the strengths of this system is that it automatically creates a dictionary linking French words to emotions.