Big Techday 12

In 2013, everything started with the talk „Programmer Anarchy” by Fred George. Is it possible to split large systems into smaller modules and to deploy them independently? Which technologies would be necessary? Which skills would our team need? Would there be only one team? How could we elude our operations-colleagues? I just ask myself why we want microservices anyway and what would happen if we actually roll it out. The network would be stable, wouldn’t it? Do I need to start with ‘Monolith First’ or with self-contained systems? Are microservices able to move dusty modernisation projects from BigBang to Build-Measure-Learn and Analyse-Evaluate-Improve while also leading to continuous improvement? In this talk, I want to share my experiences, war stories and learnings I gathered through several microservices projects, trainings and talks during the last 5 years, true to the motto “hindsight is easier than foresight”. We will talk about technologies from VMware to AWS and Kubernetes and from Netflix Hystrix to Spring Boot. We will discuss implicit assumptions, generic mistakes, incidents and what “you build it, you run it” has in common with eye bags. I also will introduce a microservice taxonomy that has been developed over the course of time. Furthermore, I will tell you about innovation tokens that helped me as an architect to not overwhelm people with too many innovations. We will also talk about the most important aspects of architecture, people and communication as well as my assumption that “it depends” might be the right answer after all.

Advent of Code is an Advent calendar of small programming puzzles for a variety of skill sets and skill levels that can be solved in any programming language you like. People use them as a speed contest, interview prep, company trainings, university coursework, practice problems, or to challenge each other. In this talk, the creator of Advent of Code will give a behind-the-scenes look at what it takes to run a month-long programming event for over 200,000 people.

Since its foundation in 2011, Celonis has grown from a small start-up to a globally distributed organization with a wide products and solutions portfolio. Not only the organization was further developed but also the product portfolio – from classic on-premise software to a Cloud-Native SaaS offering. Martin Klenk talks about the challenges on this journey and the approach that Celonis has chosen to develop a scalable and agile product portfolio.

The more people's work shifts from tactile to digital work, the more important becomes the design of surfaces for the interaction with the digital. In different application areas, we have struggled with two questions over the past few years: How can work be transferred into an interaction model in which our present work processes can be recognized, but in which helpful new digital features make a real improvement? How can the design process for such interaction models be integrated into agile software development so that a well-usable system with a permanently flexible and adaptable software architecture originates? We have found answers at various levels that we would like to introduce to the audience in this talk.

Increasingly fast development cycles, agile methods, DevOps and Cloud pose new challenges for IT systems, but at the same time create unforeseen opportunities. Leading IT companies like Amazon, Netflix, or Google are deploying hundreds of times a day in production environments. These systems can respond flexibly to short-term peak loads, attacks, and incidents and operate above 99.99% uptime. This talk will explain what distinguishes these novel resilient architectures from classic IT landscapes. Which technical and organizational requirements are necessary to make IT more resilient? How do enterprise architects contribute to resilience and how does it ultimately benefit the business?

Many large enterprises are feeling pressure: digital disruptors attack with brand-new business models and no legacy; the “FaceBook generation” has dramatically increased user expectations; and access to state-of-the-art technologies has been democratized by cloud providers. This is tough stuff for enterprises that have been, and still are, very successful, but are built around traditional technology and organizational structures. “Turning the tanker”, as the need to transform is often described, has become a board room-level topic in many traditional enterprises. Chief IT Architects and CTOs play a key role in such a digital transformation endeavor. They combine the technical, communication, and organizational skills to create business value from a tech stack refresh, to look behind buzzwords like “agile” and “DevOps”, and to build a technology platform that assures quality while moving faster. They do so by riding the “Architect Elevator” from the penthouse, where the business strategy is set, to the engine room, where the enabling technology is implemented. I rode that elevator for 5 years in a major financial services organization and am now advising major corporations on their digital journey. I collect stories from the daily life of IT transformation and package them in lighthearted, but meaningful anecdotes.

It used to be easier: the godmother called the turntable phone and announced that she had carried the package for the children to the post office. Usually, a few days later, they found a small yellow card in their mailbox noting that they could pick up the package at the branch office from 10:00 am on the next working day. Today, the online shop sends a push message to your smartphone that your package will be delivered tomorrow between 12:30 and 12:37 pm. If you are not at home then you can spontaneously redirect it to the next packet box. You can see online where the package is right now and if you want to return your purchase, simply print out the parcel label at home. Until the beginning of 2018, there were many different systems for this within DHL. All of them had their own websites, looked differently and offered diverging services. To make things worse, all systems were developed by different teams spread across Germany. Our goal was to realize a single website for DHL customers in which all services are brought together in a uniform presentation. At the same time no central monolith should be created, but instead the autonomy of the individual systems was to be preserved. In this talk, this endeavor will be considered from the point of view of a software architect. Based on three examples, I will show how an architect spends his day: coordinating developers, developing technical concepts, and ensuring that the hardware plays along.

An AI Prototype that Plays the Schafkopf Card Game: The talk gives an overview of the techniques of Libratus and AlphaGo. These include Monte Carlo Tree Search, Neural Networks, and Counterfactual Regret Minimization. In addition, it is examined whether or how they can be used in the imperfect information and multiplayer setting of Schafkopf.
Alex Sedlmayr is a Software Consultant at TNG and has studied Mathematics at the LMU Munich. He deals with Game Theory and AI and more recently with Domain Driven Design and Event Sourcing.

AI/LegalTech: The NDA Checker: Working closely with one of Germany's top five law firms, TNG developed the NDA Checker as an AI prototype for automated non-disclosure agreement (NDA) analysis and validation. The core of the application is the text analysis module developed by TNG. The application can identify certain legal clauses in NDAs. Based on this analysis result, the NDA Checker fully automatically evaluates the NDA and makes suggestions for any changes.

In order to ensure that aircrafts fulfill high quality requirements, a lot of effort must be invested in the control of component of commercial aircrafts. Artificial intelligence algorithms have great potential to automate some of the testing and streamline the quality assurance process. The example of an automated evaluation of ultrasonic test data shows how material and production errors in large aircraft fuselage shells are automatically detected with the aid of modern Convolutional Networks. An essential challenge here is to achieve a good predictive quality with comparatively few training examples.

Fraud Detection in the SWIFT-Interbanking: Especially in the area of international payment transactions, fraud is not an unusual event. In collaboration with its FinTech customer Bellin, TNG has developed a Machine Learning-based methodology that makes fraud detection easier. Therewith, payment transactions can be pre-sorted according to their fraud probability. Therefore, one must no longer examine all payments and can focus on high-risk remittances. In this talk, we introduce the technical aspects of this automated fraud detection methodology.

TensorFlow.js can be used to develop machine learning models that run on Node.js or in the browser. Existing models can also be reused or retrained. Using WebGL, the framework provides vendor independent support for hardware acceleration and can even outperform CPU-bound training. By example, I will demonstrate how to use low-level APIs and how to build a gesture classifier gathering training data from the webcam.

This session will share our experiences in converting a small remote controlled car into an autonomous driving vehicle. We'll talk about electronics, sensors, AI, computer vision and of course, the software that ties everything together. We'll introduce you into the world of self driving cars and compare our solution to the stuff that is done in the big leagues by the likes of Tesla's 'autopilot' and Waymo's self driving cars. We'll explain the challenges that have to be faced and the dilemma's that come with creating a car being driven by software in real world scenarios.

About two years ago, chatbots seemed to be the next big thing since mobile apps. In the meantime, things have cooled down a lot, with chatbots failing to deliver on the expectations. However, conversational AI is still moving forward in great strides. So, how can companies avoid the chatbot bubble and still achieve impact with the latest conversational technology?

In this talk, I will give an introduction of Deep learning in the medical field, and in particular its application to the processing of medical images and genetic data. Medical Data Analytics is a field of Data Science which has its own specificities and challenges, such as data scarceness and variability, and where the black-box aspect is a big drawback. As a use case, we will present the work carried out at Konica Minolta and the TU Munich: starting from raw medical scans and patient data, we detect pathologies and extract relevant features automatically, in order to derive models and insight of medical interest.

Imagine that you are standing in front of a mirror, but no longer see your own face, but through the eyes of Barack Obama or Angela Merkel. In real time, your own facial expressions are transferred to someone else’s face. The TNG Hardware Hacking Team has managed to create such a prototype and transfer a person's face to any other face in real time. The basis for this is the so-called deep-fake approach. The use of neural networks detects faces of video input, translates and integrates them back to the video output. Through this technique, it is possible to project deceptively real imitations to other people. For this purpose, we used Keras trained autoencoder networks and various face recognition algorithms. In this talk, Thomas Endres and Martin Förtsch give one entertaining and very vivid introduction to the world of deep fakes in real time. In doing so, they particularly focus on deep learning techniques used in this application.

While recent approaches based on neural networks result in significant improvements in the quality of machine translations, human translators are playing an increasingly important role in introducing these technologies. In this lecture you will learn about the basics of machine translation systems. We examine the question which parameters promise the best translation quality and what is important when using machine translation in order to realize enormous efficiency gains in the professional translation environment.

In December 2017 Google DeepMind published a paper about AlphaZero, a new chess engine of unique style and unprecedented strength. Unlike traditional engines, AlphaZero did not use any human chess expertise and learned exclusively through self-play. However, DeepMind did not release any source code, so chess fans outside of DeepMind could not benefit from that novel idea. Leela Chess Zero is a faithful open source implementation of the ideas described in DeepMind's paper. In this talk I'll describe how Leela Chess Zero became the strongest chess engine available after just one year of the project's existence, what the differences are from traditional chess engines, and lessons learned for running a distributed open source project.

Facial analysis technologies have recently surpassed the capabilities of expert clinicians in syndromic phenotypes identification. To date, these technologies could only identify phenotypes of a few diseases, limiting their role in clinical settings, where hundreds of diagnoses must be considered. DeepGestalt uses computer vision and deep learning algorithms to highlight numerous genetic syndromes correlating with patients’ phenotypes analysed from unconstrained 2D images. DeepGestalt achieves 91% top-10-accuracy in identifying over 200 different genetic syndromes and has outperformed clinical experts in three separate experiments. We suggest that this form of artificial intelligence is ready to support genetics in Clinic, Research and Variant Analysis practices and will play a key role in the future of precision medicine. In this talk I will review DeepGestalt technology and will demonstrate its use in each aspect.

Biometric Identification Systems are especially useful in situations where people do not have identity documents or cannot show them. One possible application is a fraud prevention system for charitable services which would need to be portable and run on low end commodity hardware like mobile phones. During the talk we will present classical as well as deep learning approaches to tackle the challenge of learning similarity metrics suitable for identification and show how open source software can be used to build prototypes within a very short amount of time.

Chess and mathematics belong to the intellectual world cultural heritage. Since its creation, chess has been played throughout the world in almost all cultures. Mathematics is a human resource that has grown over thousands of years. It is – often unnoticed – in many things that surround us. The heating heats, the plane flies only when math is involved. Chess and mathematics are both sources of sustained perceptible beauty. Here and there the aesthetics are in the radiance of cleverly linked ideas. The talk shows highlights of these two worlds of ideas as well as the manifold relations between chess and mathematics.

The Early Days of Id Software: As co-founders of id Software, John Romero and John Carmack created the code behind the company's seminal titles. The principles they defined through experience in id’s earliest days built upon one another to produce a unique methodology and a constantly shippable codebase. In this talk, John Romero discusses id software’s early days, the programming principles behind it and the events that led to their creation.

The history of women in computing has largely been lost, like the histories of factory workers who built the first cars. Yet, women invented programming, were the original developers for the ENIAC, created assembly language and developed the first compiler (not to mention the term “compiler” and “bug”), and were instrumental to the development of many seminal programming languages. So what happened? It’s a drama that’s equal parts cultural excavation and celebration. In this talk, Brenda Romero digs up this fascinating history, explores what happened, and looks at how the artifacts of this legacy still affect computing and its growth today.

In 1968, a Soviet nuclear ballistic missile submarine went missing in the remote Pacific and the Russians declared it lost. Under the veil of secrecy, the Americans found the wreck, and over the next 6 years, the CIA designed and executed perhaps the largest and most complicated covert spy mission in history. This mission, code-named Project Azorian, was beyond audacious. Engineers set out to build a ship and system that could pull a 2 million pound object off the floor of the sea, 16,500 feet under the surface, without anyone knowing its true purpose. How is that possible? Because the CIA had the perfect cover story. The ship, they told the world, was an ocean mining vessel with a very eccentric owner: Howard Hughes.

Space applications are a crucial element in today's digital economy, enabling global communication networks, logistics monitoring or providing data for business analytics. With commercially-driven companies entering the market, the future is looking bright on a global scale. How can we harness space data and who is taking the lead in the space race 2.0?

BepiColombo is a space mission to Mercury, developed by the European Space Agency ESA in collaboration with the Japan Aerospace Exploration Agency (JAXA). The mission is composed of two scientific orbiters, which were launched together on 20th October 2018 as a single composite spacecraft, including a module with electric propulsion to support the 7-years cruise phase which includes also planetary swingbys at Earth (1x), Venus (2x) and Mercury (6x). The spacecraft is operated from ESA's European Space Operations Center (ESOC) based in Darmstadt, Germany. The presentation will introduce the mission scientific objectives and specific challenges, as well as report on the main events since launch.

How can we realize a multiplayer game of volleyball in VR? Is that even possible? Which pitfalls will we encounter on the way? And will the game be fun? Today we can give answers to these questions that drove us a year ago. During this talk, we will report on the challenges that we had to overcome on our way to a completed game. Due to the advent and easy access to professional gaming engines like Unity 3D, Unreal and Cry Engine in recent years, programming games has become much easier. Nevertheless, the specific challenges during the implementation of individual games remain. We will especially focus on the difficulties of dealing with network latency in conjunction with the physics engine. Also, we will show how we worked iteratively on the mechanics of the game and the physics of the ball in order to achieve not only a seemingly realistic but - most importantly - entertaining game behaviour.

Disney Research was launched in 2008 as an informal network of research laboratories that collaborate closely with academic institutions such as the Swiss Federal Institute of Technology in Zurich and Carnegie Mellon University. Its mission is to push the frontiers of  technology in areas relevant to Disney's creative entertainment businesses. Disney Research develops innovations for Parks, Film, Animation, Television, Games, and Consumer Products. Research areas include video and animation technologies, postproduction and special effects, digital fabrication, robotics, and much more. This event gives an overview of Disney Research spiced with some examples of our latest and greatest inventions. We will focus onto the collaboration between ETH Zurich and the Walt Disney Company and display the synergies arising from this program. The two presentations will highlight a company perspective as well as a view from the academic angle and they will be followed by a panel discussion.

“It is all about generating added value to products and economic value to the customer” – Product examples that used additive manufacturing show various added values. Weight savings of up to 50% in the aerospace industry save operating costs and kerosene. Homogenous heat transfer in heat exchangers increase efficiency and performance. Volume reduction of 30% in tooling due to functional integration shortens the supply chain. Furthermore, conformal cooling or heating enables a higher efficiency of tooling and shortens the production time. Economically valuable serial products are only made possible by the perfectly harmonized triad of design, materials and production as APWORKS offers it. Design means developing products especially for additive manufacturing. That means considering the conditions for subsequent process steps from the start of the design process. Combining the benefits of metallic 3D printing with new materials can greatly expand the possibilities for modern components. Scalmalloy® is such a best practice example, as it combines light weight of aluminium with the specific strength of Titanium. The combination of design freedom of additive manufacturing and a high-performance material permits the execution of highly complex geometries. Accompanying the advanced design and material capabilities the production facilities need to feature latest additive manufacturing equipment.

Wind turbines, sometimes called WECs for wind energy converters, are nearing a total weight of 1,000 tons and represent the result of modern mechanical engineering wizardry. However, many operators of wind farms face various types of curtailment; even right after launch, parks often deliver significantly less energy than what was expected. Operational maintenance continues to be done in fixed intervals. Towards the end of the certified lifetime, the economic viability of continuing operation and its alternatives needs to be quantified, but data is hard to come by. And this is a recurring theme in the wind industry: (non-) availability of data and the questionable quality and resolution of the data that is. The presentation will take you through a handful of case studies of how fiber-optic sensors are integrated into the rotor blades to provide high-frequency (in this context, 40 Hz) data from which we calculate various relevant physical parameters (wind direction, speed, sheer (vertical and horizontal), turbulence, etc.), but can also assess the condition of the blade (icing, crack, lightning strike, etc.). We do all this to increase the annual energy production (AEP), cut operating expenses, and prolong lifetime. And yes, the effects are highly measurable. We have been at this since 2012 and the technology is protected by more than 100 patents. Looking forward, our beliefs are that owners of assets will use data driven optimization and that the turbines themselves will be online and interacting. Assets will go from underused to understood, and we will do our share to make this a reality faster.

Quantum computing promises to revolutionize computer technology as profoundly as the airplane revolutionized transportation. After decades of incubation, early generation quantum computers are finally appearing that allow people to begin experimentation in earnest. In this talk, I will describe D-Wave's approach to quantum computing, explain its pros and cons with respect to competing schemes, and give the rationale behind our design choices. Furthermore, I will give examples of how the native optimization and sampling capabilities of our quantum processor can be exploited to tackle problems in a variety of fields including healthcare, physics, finance, simulation, artificial intelligence, and machine learning.

Computational drug discovery aims to develop algorithms for the rational design and optimization of therapeutic molecules. When studying the dynamics of drug targets as well as protein therapeutics, one needs to be able to predict their three dimensional structure from the amino acid sequence - widely known as the protein folding problem. Computational docking algorithms are used in order to rank various drug compounds according to their interaction strength with the drug target. Most approaches to protein folding and docking are computationally intractable with classical supercomputers. In this talk, I will introduce ProteinQure's portfolio of published quantum algorithms. This includes two hybrid approaches to folding coarse-grained lattice proteins on different types of quantum hardware. As part of this, I will demonstrate how we folded the largest protein on a D-Wave quantum processor. Next, I will outline how quantum computers can enhance traditional software packages since we've developed a quantum-enhanced version of Rosetta, the world's most popular protein engineering software. Lastly, I will introduce a quantum-classical hybrid graph algorithm for predicting molecular docking configurations that is optimized for photonic quantum chips which can be operated at room temperature.

One of the biggest societal issues nowadays is the ever-increasing demand for energy, dwindling fossil fuels and the drive for clean and sustainable power generation. Among the most common renewable energy technologies, solar power generation has perhaps the greatest potential due to the immense energy production of the sun. Until now silicon-based solar cells are ubiquitous, but they are far from being sufficient to cover the entire energy needs of the earth. In the search for new materials to improve or completely replace silicon solar cells, the so-called halide perowskite appeared in 2011 for the first time. Due to their extremely advantageous optical properties, perovskite-based solar cells have since been improved so far that peak values of the conversion efficiency already come close to the silicon-based solar cells. In this lecture the material of the halide perowskite is presented and its fundamental properties and the possible optoelectronic applications (e.g. solar cells, LEDs) are explained. In addition, it addresses the (still) existing problems that currently prevent the wholesale commercialization of this highly interesting material.

Digital transformation represents an opportunity for improving productivity growth by enabling innovation and reducing the costs of a range of business processes. Yet despite the rapid advance of digital technologies, aggregate productivity growth has slowed over the past decade or so, raising the question of how digital technologies can boost productivity. Today, as in the 1980s, when Nobel-prize winner Robert Solow famously quipped: "we see computers everywhere but in the productivity statistics" there is again a paradox of rapid technological change and slow productivity growth. OECD work shows there is hope for the future. While not yet showing up in the aggregate productivity data, digital transformation is starting to have impacts on productivity in individual firms – and increasingly also in certain industries. Further and larger impacts should emerge as digital transformation evolves, especially in the wake of Artificial Intelligence, and as digital technologies, business models and practices diffuse to a greater number of firms and industries. Policy makers can help ensure that these impacts emerge by engaging in supportive policy actions, in particular for less productive firms. This would result in a double dividend in terms of productivity outcomes and inclusiveness.

With ever-increasing amounts of data and real-time requirements, the cloud often no longer provides a solution for IoT applications or mobile applications. The trend is therefore (back) to edge computing and autonomous programs. What is Edge Computing? What are advantages and disadvantages? Which practical examples are there? What is happening in edge computing right now? And what does ObjectBox have to do with it? These questions are discussed in the speech.

Once upon a time, we used software that ran on our own computers, that worked offline, and that stored its data in files on the local disk. Then we decided to put it all in the cloud. We gained some great features: real-time collaboration, like in Google Docs, for example. But we also lost control of our own data, and became dependent on far-away servers to allow us to access the data that we created. Automerge is part of an effort to get the best of both worlds. It is a JavaScript library for building real-time collaborative applications. However, apps built with Automerge also work offline, storing data locally, and synchronise their data with collaborators whenever a network is available. And although you can use it with servers, you don’t have to: synchronisation also works peer-to-peer, or via any network you choose. In this talk we will explore how Automerge deals with different users independently modifying shared data in a collaborative application (hint: by merging the changes… automatically!), how it achieves consistency in highly distributed settings, and where it is heading in the future.

Big test suites often have a long run-time. Therefore, in practice, they are often not performed as part of Continuous Integration (CI), but only in later test phases. Unfortunately, many errors remain unrecognized during the CI and are located too late, causing high costs. Test Impact Analysis allows you to run only those tests that have been affected by the code changes since the last test run and are, therefore, most likely to contain new bugs. In our empirical samplings, we were able to find 90% of the errors in 2% of the test execution time. The talk presents basics, research results, and empirical samplings on test impact analysis. In addition, we report on our experiences in our own development process, in open-source projects and with customers.

In search for a feature-complete solution that supports HA clusters, follows Kubernetes best-practices, and comes with a simple and declarative API based on the Kubernetes Cluster API, we could not find a single one that fulfilled our needs. Therefore, we decided to build
KubeOne. KubeOne takes care of installing, configuring, upgrading and maintaining Highly-Available Kubernetes clusters. It works out-of-the-box on any cloud provider, as well as on-prem and in bare-metal environments. In this talk,  I will introduce KubeOne and show how to use KubeOne to:

• Set-up HA K8s clusters on any cloud provider

• Provision, upgrade or destroy the cluster with a single command

• Configure features such as PodSecurityPolicy or DynamicAuditLog

While Docker is still most popular for building and running containers, it has security and scalability shortcomings for production systems and build pipelines. Recently alternatives have emerged to build container images without Docker. Each of these address common problems: building without elevated privileges, reproducible results, caching of intermediate layers and scaling CI/CD in larger organizations. We first introduce the basic structure of a container image and compare the build process for a selection of these tools. We further demonstrate their usage and discuss strengths and weaknesses of each. Finally, we give guidance for selecting the right tool – which might not always be Docker.

ArchUnit - Unit Test your Java Architecture: This presentation will give a quick introduction to the Open Source library ArchUnit. ArchUnit allows to bring architecture and (JVM) code closely together by executing architecture rules as plain unit tests with any established test framework. This helps in particular with big agile projects where the role of the software architect might be distributed and business goals are adjusted iteratively with a high frequency. ArchUnit is very easy to introduce, free to use and can be established on a minimum budget. The talk will line out some of the core features and how ArchUnit can help in any non-trivial software project to achieve a maintainable and consistent code base, which is furthermore automatically in sync with architecture documentation.

Everyone is talking about DevOps - but how do you do it? Google has published its internal best practices under the name Site Reliability Engineering. We took an intensive look at this subject. In this lecture, we present the key concepts and show applications from our everyday project life. If you've always wanted to know what's behind the terms Infrastructure as Code, Error Budgets, or Service Level Objectives, then this is the right place for you. Technology should also not be neglected - the listener learns which tools and frameworks can support a reliable and secure cloud operation. Cloud-native technologies such as Kubernetes and Prometheus play a central role, as do CI / CD tools such as GitLab or Terraform.

Most people agree on how important an interdisciplinary and entrepreneurial mindset is to succeed in Information Technology. However, the sad reality in many companies and educational programs is still a deep divide between “business” and “technical” people and matters. My talk will cover why this needs to change and why I truly believe that “technical” people are best equipped to add tremendous value to their companies. I will share practical advice on how students, young professionals, and their managers can get there.

A new bug was found in a well-known virtualization software, for which there was no, at least not publicly known, working exploit. The purpose of this example is to demonstrate how a small bug in a software can override the overall security of an application and the underlying operating system. Also we will see how a real exploit is developed and implemented. In addition, the assumption "virtualization equals sandbox" will be contended.