Search
A virtual representation of a brain hovers above a robot arm.

Artificial intelligence from the Open Telekom Cloud

The Open Telekom Cloud provides you with scalable, secure and privacy-compliant computing capacity for your AI project

How AI projects succeed

Would you like to try out initial prototypes and applications with artificial intelligence (AI), but are lacking the necessary capacities and competences in your company? And you also shy away from high investment costs? Thanks to cost-effective public cloud capacities, artificial intelligence has become affordable for every company. Use resources from the Open Telekom Cloud easily, flexibly and quickly and start your AI project!

The A.I.-Team | Artificial Intelligence Expertise

T-Systems has already developed, implemented and supported countless AI projects for companies. Our AI experts report from their experience and provide exciting insights into questions such as: What is an authentic AI? Which technology is the best? How do AI projects succeed? Everything worth knowing in our videos!

Cloud capacities for powerful AI operation

Whether digital speech assistants, precise predictions of customer behaviour in marketing, or the predictive maintenance of machines in Industry 4.0 – the applications for artificial intelligence are manifold. Intelligent algorithms evaluate extremely large amounts of data from a wide variety of sources, for example, through machine learning or neural networks. They also recognise patterns and connections that people tend to overlook. They produce accurate forecasts, are capable of learning, and develop independently. However, this only works with a powerful infrastructure, e.g. for training algorithms with large amounts of data. This is where companies benefit from a clever combination of on-premises capacities and demand-based IT resources from the public cloud. And at processor level, from the needs-based use of CPU, GPU, FPGA, and the like.

Processors for AI: speed is not sorcery

FPGAs and artificial intelligence

Artificial intelligence applications often require real-time responses with the lowest possible latency – this includes online shops that update prices or trading platforms that need to react to market developments as quickly as possible, as well as short-term weather forecasts for storm warnings. If there is no time to lose, field programmable gate arrays (FPGAs) from the Open Telekom Cloud are the best solution.

Programming FPGAs individually

FPGAs are programmable hardware elements or circuits. In contrast to processors such as CPUs and GPUs, the sequence and type of use of chips, memory, networks and the like are not fixed. The circuits are freely configurable, very flexible thanks to their reprogrammability, and are used in the context of AI for machine learning, fraud detection, speech recognition and trend analysis, among other things.

Turbo-boost for applications: Xelera

The startup Xelera demonstrates how FPGAs can be used as accelerators for corporate applications with its suite, which you can book from the Open Telekom Cloud. The Xelera Suite is a hardware-independent interface to FPGA platforms in data centres and the cloud. It allows processes and applications in companies to be brought up to speed without specific technological know-how: depending on the scale, accelerations of more than 100 times are possible.

The capacities from the Open Telekom Cloud allow us to set up our accelerator quickly in the cloud and configure it according to the customer’s application scenario, for example, accelerating standard databases such as SAP.

Dr. Felix Winterstein, founder at Xelera

Open Telekom Cloud test winner in the analyst benchmark

The analyst firm Cloud Spectator tested highly specialised virtual machines (VMs) from various cloud service providers, including Amazon Web Services (AWS), Microsoft Azure, Google Compute Engine (GCE) and the Open Telekom Cloud for the benchmark “Western Europe Cloud IaaS Analysis.”

Two types of virtual machines (VM) were tested: firstly, high-performance VMs with non-volatile memory express (NVMe) or local SSD memory, and secondly, memory-optimised VMs with correspondingly high read and write performance. All of these VMs meet the necessary requirements for specific applications, such as processes requiring extremely high data throughput (IOPS), floating point operations, real-time analysis, processing large amounts of data, or performing high-frequency database operations in random access memory (RAM).

The conclusion: “Cloud Spectator’s analysis showed that the Open Telekom Cloud portfolio from T-Systems offers premium solutions that are superior to comparable hyperscaler offerings. Both of the tested VM types from Telekom have significant advantages over their competitors,” said Cloud Spectator when summarising the results of the study.

AI in practice

AI has all the prerequisites to completely renew companies from head to toe. Nearly all industries and disciplines can benefit from artificial intelligence applications. Find out how you can benefit from AI in practice here.

AI in the retail sector

Personalised marketing in the retail sector

Using image processing and pattern recognition, intelligent systems on digital screens in salesrooms can help to tailor advertising and the way in which customers are addressed to make personalised offers. To do this, the AI systems evaluate characteristics such as posture and viewing times for specific goods displays. If several networked devices are used, they learn from each other and increase AI effectiveness.

Intelligent audio marketing in shops

The selection of background music influences the shopping behaviour of customers. AI can be used to identify the number and age of people in a store and link them to business data such as inventory levels and external influences such as weather and local events. Based on this analysis, background music and promotional ads are selected to create a pleasant shopping experience for customers.

Predicting customer behaviour

Learning from customer data and predicting customer behaviour

Machine learning (ML) processes can be used to prepare existing customer data in such a way that they enable forecasts of purchasing behaviour. Factors such as the time spent on a website, number and prices of products in shopping baskets, age of customers and order histories are evaluated. The ML system can independently derive patterns and predictions from this information – for example, that cheaper products will be purchased faster, that repeated call-ups on a product will increase the probability of an order, or that certain products will be purchased frequently by customers in a certain age group.

Tracking down trends and customer preferences in social networks

The use of AI processes such as machine learning is not limited to the company’s own data stocks. AI can also be used to analyse posts and tweets on Facebook or Twitter, including information about word frequency and emotional content (sentiments) as well as the number of hashtags and likes. In addition, the systems can take weather and news situations into account in their analyses and incorporate them into forecasts of future customer behaviour. Machine learning enables AI systems to independently recognise and evaluate relevant features and threshold values. And artificial intelligence never stops at the status quo – it continues to learn and develop.

AI in Quality Assurance

Quality assurance in production: article inspection

A self-learning surface inspection ensures that surface defects, contamination and variations in article patterns are reliably detected. The Fraunhofer IPA has developed an adaptive, optical inspection method for this purpose. It is based on an unsupervised learning process in which the optical inspection system automatically adapts to changing surface structures.

Quality assurance in industry: recycling

With the help of a self-learning object classification, waste equipment such as catalytic converters can be better identified. Their condition is often poor due to corrosion, deformation or damage. The devices are recorded three-dimensionally with a laser line sensor and classified on the basis of neural networks using various characteristics – such as object contour and curvature. Fraunhofer researchers achieved a classification rate of over 90 percent.

Quality assurance in the pharmaceutical industry: visual inspection

Large quantities of fertilised zebrafish eggs are required in pharmacy and biology. Thanks to an automated visual inspection, fertilised eggs can be automatically recognised and distinguished from unfertilised ones. To do this, a camera image is taken of each egg and analysed by a deep learning network. The detection rate is 99.8 percent.

Further practical examples

Smart wearables – keep an eye on everything with data glasses

Portable, sensor-equipped AI systems that are worn on the body or integrated into clothing provide useful services. Data glasses, for example, support employees in picking orders by scanning containers, warehouse and article codes. For example, coloured markings on the glasses can make it easier to find the right shelf or maintenance instructions can be brought up. This saves time and minimises the error rate.

Control vehicles from over 4000 kilometres away

Together with the Israeli start-up Ottopia, T-Systems has shown how vehicles can be remote-controlled, transferred or even used as taxis in a factory yard from thousands of kilometres away. To ensure that the response time for transmitting the huge amounts of data is not too long, Ottopia uses an AI application that predicts the utilisation of the radio cells in good time. This enables uninterrupted services, even over public LTE networks in remote locations.

AI-supported route planning reduces empty runs

One in three trucks drives through Germany empty. AI-supported route planning, which takes into account factors such as weather, events and customer buying behaviour at different times of the year, can reduce the number of empty runs and standing times. The AI system forecasts the market prices for truck routes and determines prices for spare capacities. In this way, supply and demand can be coupled quickly and effectively – and transport journeys made more sustainable.

AI sorts the mail – automated document capture

Letters, e-mails, invoices and other documents can be automatically processed and distributed using the image and object recognition of AI solutions. Since documents often arrive at companies in different formats – especially companies that are active internationally – it is advisable to use a self-learning AI solution that can correctly assign text, numbers and values.

Understanding texts correctly – with text mining

Text mining is an AI-supported reading and comprehension aid that uses methods such as natural language processing and deep learning. It recognises and processes contextual meaning in structured and unstructured documents. Numerous processes benefit from this: for example, the analysis of customer feedback in e-mails can be used to automatically trigger the right response from customer service. In addition, text mining is very useful when it comes to the analysis of contract texts and other business documents.

Curious?

Our AI experts determine the potential your company has to benefit from artificial intelligence. We work with you to implement individual AI projects – from concept to installation and operation in the Open Telekom Cloud.

Do you visit t-systems.com outside of Indonesia? Visit the local website for more information and offers for your country.