Artificial intelligence and machine learning are hot topics in almost every domain. Also in software development the technologies can be used to assist developers to be more efficient, and for users to get software more tailored to their needs. In software, the methods come with different names, such as data-driven software development and continuous experimentation. How can we get the machines to do the jobs for us?

Bio

Charlotta has 20 years experience from various management positions in IT and telecom focusing on international business development and organisational development. She has served as COO at Axis and CEO of Decuma and TAT The Astonishing Tribe. Since 2003 she has been board director of various public and private companies, primarily within the tech sector, and since 2011 works full time as board director, currently serving at Bure, Invisio, Net Insight and CLX. She is chairman of the board at the Faculty of Engineering at Lund University and engaged in the start-up community of southern Sweden. She has a master’s degree of business administration and an honorary doctorate degree in engineering.

Martin Monperrus, Professor of Software Technology, KTH Royal Institute of Technology

Imagine a robot automatically fixing your Travis build failure. Instead of understanding the Travis log, and fixing the error, you would merge a pull request stating ”Fixing build failure #1234”. This talk shows that this is possible. It starts with an overview of academic research on automatic repair, and then follows up by stating the vision of automatic repair in continuous integration. Can this be done for real? I’ll present you Repairnator, a robot who automatically fixes build failures, and the scientific and engineering challenges behind such a robot.

More about Martin Monperrus.

Christoph Reichenbach, Senior Lecturer, Lund University / LTH / Computer Science

Software developers often develop new code by copying, pasting, and modifying code that is already similar to what they need. However, the research literature shows that this approach increases the maintenance cost later on– by copying and pasting, developers trade off quick results now for extra work later. Conversely, more conscientious developers need to work longer and harder than their copying and pasting colleagues to avoid this extra work down the road.

We have developed an approach that presents a way out of this dilemma: Copy-paste-modify-merge, a technique in which developers copy, paste, and modify as before, but then invoke a specialised refactoring tool that merges their modified copies with the original. The result is a single piece of code that is parameterised to allow all the functionality that each of clones offered, plus easy-to-use wrappers that simplify the interface.

Our initial research results show that our approach is effective at producing merged code that is good enough to meet the quality criteria of several industrial Open Source projects.

David Mattos, PhD candidate, Chalmers University of Technology

Customer behavior data, machine learning and online field experiments are extensively used by software companies to optimize their systems, deliver value to their customers and provide a competitive edge in their business. Optimization of systems parameters can be both a challenging task when the behavior is stochastic (such as customer behavior), and a tedious task when the expected behavior is known. The former case is traditionally done using sequential experiments (multiple A/B/n testing), while the latter traditionally uses specialized optimization software and search methods (grid search, simulated annealing and random search). Up to date, there is no software that can integrate both approaches in a unified and automated way. We will present a machine learning software system using state-of-the-art reinforcement learning algorithms that can integrate both approaches. In this talk, the system is presented together with simulations and a case study with Sony Mobile, showing how the system can be integrated while keeping minimum coupling with the application code.

Rasmus Ros, PhD candidate, Lund University
Mikael Hammar, Head of Research, Apptus Technologies AB

What are the correct settings for an optimal user experience? Controlled experiments, or A/B testing, has been used to answer this for over a decade to choose feasible settings. But what happens when the answer changes over time? Or if there are hundreds of variables that interact in intricate ways? Or if the answer must be personalized for different users? A/B testing cannot handle the necessary speed in decision making nor the scale of such problems.

The solution to this problem is machine learning or genetic algorithms that continuously learns from users as the software is being used (known as reinforcement learning). We present a framework and code examples to show how to set it up in practice. We emphasize how developers can control and guide the algorithm from making unwanted decisions (using constraints). The framework has been evaluated in an industry-academia collaboration with Apptus in Lund. Apptus develops automatic optimization solutions for e-commerce web shops around the world. They provide a unified solution for product search, recommendations, navigation, and more.

When Open Source first started it was called Free Software and had its roots in ideology and far from corporate structures.  Over the last 20 years, the term open source was created and open source software(OSS) has become a platform for innovation for companies.  Almost all proprietary products today have some amount of OSS and increasingly companies are collaborating with each other on non-differentiating software.  The use of OSS also introduces a very different way of development and a culture change inside companies who now need to balance open source collaboration with corporate value creation and structures.  I will explore some of these changes that companies have to embrace and how they balance value creation with collaboration and community engagement.

Bio

Nithya A. Ruff is the Senior Director for Comcast’s Open Source Practice. She is responsible for growing Open Source culture inside of Comcast and engagement with external communities. Prior to this, she started and grew the Western Digital’s Open Source Strategy Office. She first glimpsed the power of open source while at SGI in the 90s and has been building bridges between companies and the open source community ever since. She’s also held leadership positions at Wind River (an Intel Company), Synopsys, Avaya, Tripwire and Eastman Kodak. At Wind River, she led a team of product managers in managing a world class embedded Linux distribution and was a key member of the Yocto Project advocacy team on the board. Nithya is a Director at Large on the Linux Foundation Board and represents community interests on the board.

Nithya has been a passionate advocate and a speaker for opening doors to new people in Open Source for many years. She has also been a promoter of valuing diverse ways of contributing to open source such as in marketing, legal and community. You can often find her on social media promoting dialogue on diversity and open source. She has spoken at multiple conferences such as OSSummit, OSCON, All Things Open, SCALE, Grace Hopper, OpenStack, VMWorld, OS Strategy Summit and Red Hat Summit on the business and community of open source. In recognition of her work in open source both on the business and community side, she was named to CIO magazine’s most influential women in open source list. She was recently one of 4 people to win the 2017 O’Reilly Open Source Award for exceptional contribution to open source.

Nithya graduated with an MS in Computer Science from NDSU and an MBA from the University of Rochester, Simon Business School. She lives in the San Francisco Bay Area and is a proud mother of two daughters. You can follow her on twitter @nithyaruff

Johan Jönsson, Technical Collaboration team, Wikimedia Foundation

Wikipedia is the world’s most popular open-source website. It’s owned and operated by a non-profit foundation, the Wikimedia Foundation, but all content is created on a volunteer basis and the administration and curation fall on the editor community.

They want a say in the technical development. As do others. Wikipedia runs on MediaWiki, wiki software that’s become very popular for wikis outside of the Wikipedia sphere as well.

The MediaWiki software development is done by Wikimedia Foundation staff, staff at affiliate organisations who support Wikipedia in their part of the world, volunteer developers who care about Wikipedia and third-party developers who want to tailor it to their needs outside of Wikipedia. On Wikipedia, some users can edit parts of the site’s CSS and JavaScript and yet others run bots that interact with the wikis.

With all these voices, how do we make decisions? What are the challenges when affiliate organisations decide to work on a project? How do salaried and volunteer developers work side by side? How much attention do we pay to the wishes of third-party requests who want to use the MediaWiki software outside of our wikis?

A brief presentation of successes and failures.

Nicolas Martin-Vivaldi, Consultant, Addalot Consulting AB
Carl-Eric Mols, Head of Open Source, Sony Mobile

Open source software can provide significant benefits to an organization. Many of today’s fast-growing companies like Amazon, Google and Netflix, have embraced Open Source as part of their business strategy.

We are now seeing a new wave of development being led by established industrials that are growing with Open Source at the heart of their business. In contrast to legacy practice and locking in their IP assets, they are making vital software assets publicly free for anyone to see and use. How do they (dare to) do it?

These systems are no longer primarily build on inhouse development but on a mix of inhouse/ third party/open source. Software companies has gone from development to integration. What capabilities are now needed?
We call this Industrial Open Source!

• Based on Sony Mobiles experiences a handbook with industry proven and standardized patterns of how to manage Industrial Open Source software in large organizations. The aim of the handbook is to guide on an Open Source Program.
  Some key takeaways:
  – Why compliance is a necessity
  – Why contribution is vital
  – How to create more business with open source

Hussan Munir, PhD candidate, Lund University

The increased use of Open Source Software (OSS) affects how software-intensive product development organizations innovate and compete, moving them towards Open Innovation (OI). Specifically, software engineering tools have the potential for OI, but require better understanding regarding what to develop internally and what to acquire from outside the organization, and how to cooperate with potential competitors. The presentation reasons that openness provides opportunities to reduce the development cost and development time. Furthermore, OI positively impacts on the process and product innovation, but it requires investment by organizations in OSS communities. By betting on openness, organizations may be able to significantly increase their competitiveness

Lothar Wengerek, Line manager, Ericsson AB
Christopher Price, Head of Ericsson Software Technology, Ericsson AB

The Ericsson Inner-source and Open-source project ””Theia – Gaia”” consists of several scalable services and web based IDE’s which are interconnected in a cloud-based infrastructure.
In this presentation we will explore those services and how they can be used to accelerate product development.
Today, product development is experiencing tremendous disruptions in the same way as we go through technology shifts.

To cope with this speed of change we also must adjust quickly the way we are developing things.
Automation and build-in machine-intelligence can be a decisive advantage to shorten time to market for our products
Experiences from current internal users are showing that the services can help to increase collaboration, efficiency, and quality.

The system is designed for inner-source and open-source projects. Specific pre-defined workspaces contain “all you need” to participate and the only prerequisite is a modern web browser to be able to study, experiment with and contribute to the project.