WeAreDevelopers Live Week
Our Senior Consultant, Linda Mohamed had the honour to share her passion of machine learning at the WeAreDevelopers conference. Furthermore, we interviewed her about her experience as a speaker and what key takeaways she got from one of the worlds most famous developer conference.
What is WAD?
"WeAreDevelopers" is a community whose mission is to provide developers with the things they need and love. It sees itself primarily as a job placement platform for developers and companies and is behind the WeAreDevelopers conferences. They try to give the developers a voice and connect them. They do it with their regular events where developers can talk about applications and frameworks, architecture and patterns, and future technologies and problems that could evolve. This year, in 2020, they held it entirely virtually for the first time - and I was invited to play a part in it.
Another reason why they intend to build a developer community is their business model of job placement - they locate vacancies from Europe's leading technology companies and place developers. In this way, they connect not only developers but also potential employers.
Additionally, Heise Medien has recently acquired a financial stake in the start-up company WeAreDevelopers.
What do you associate with WAD?
For me, the WAD is an excellent opportunity to dive into new technologies and find people who already have experience with these new technologies. As I also mentioned in my presentation, "Leverage cloud computing with serverless multi-cloud machine learning," it is easy to learn new things when you have someone to ask. The WAD creates an environment and a platform that enables precisely this and provides a voice to a target group that is otherwise not so active in sharing content in presentations and lectures to the general public. With topics such as the democratization of technologies, cloud computing, frameworks for cross-platform development, etc., they offer mainstream content for a bigger audience than just software developers. Nevertheless, they do not lack in-depth expertise - in specific technologies, concrete problems from real-world examples are addressed and explained.
Why did you decide to talk about Machine Learning?
At the beginning of November, I had a presentation within the NewITGirls community, where I talked about automation, digitization, and digital transformation. In this talk, I explained what information technology is and where it can support companies—speaking of digitization - which is just the process of compressing analog media into bits and bytes to make it available in digital form—we can also call it data. This data can be films, pictures, different kinds of information in written form, and much more. But analog and physically available data is always digitally reproduced, copied, or transmitted - do you see the problem?
Let's come briefly to the subject of IT - by definition, it means the use of computers to store, retrieve, transfer, and manipulate data or information. So when we think of supporting companies, it goes far beyond converting analog data into digital form. So the term IT is typically used in the context of business operations and transactions. To continue with the example, IT helps companies make their data available in digital form and retrieve, manipulate, and transfer it for various purposes.
With the fact that the amount of generated data within businesses is growing and growing, the workload of systems and requirements of data processing in different forms increases. As a result, required IT systems and components must be able to react more flexibly to changes in workload and process large volumes of data in a short time.
IT systems typically work with defined data as input, process this data according to business requirements, and return the result in the form of data as output. The more information we have, the more complicated will the business logic probably be.
As complexity increases, so does the processing. In a nutshell machine learning means, using computer algorithms that improve automatically through experience. Thus it is a technology that provides a significant advantage to gain meaningful insights in large amounts of data because it reduces the effort of building complex decision trees manually - which means, less coding. Well, and when AI is the concept of using historical data and human experience to solve problems, we can move this use case further towards AI. With AI, companies can use these insights to automate intelligent behavior within a more complex deep learning model to support humans in their decision-making process or even decide on behalf of humans.
These possibilities combined with the advantages of using serverless technologies and cloud computing are the reason why I decided to experiment with this technology and learn more about artificial intelligence and its underlying techniques.
What could you take away from your presentation?
Share your experience with us.
During the preparation for my presentation and the development of the prototype, which I explained within my talk, I got essential insights into current problems. Artificial Intelligence is a multidisciplinary technology because you have to combine several areas of expertise to solve just a single concrete problem. So I learned that you could use AI to solve specific issues, as far as you have enough time plus data and the needed domain knowledge for the use case. The lack of high-quality and varied data leads to the next challenge, the so-called bias, which can be described as a tendency to lean in a specific direction, either for or against a particular cause, leading to discrimination. As diversity and quality of data are essential, I tried to create a feature engineering mechanism to scale this process. I would consider this as the first step towards a deep learning model. During this process, I recognized that traceability is getting worse the more different data sources and use cases you are trying to include in your model training. If, on the one hand, you need to ensure that decisions are traceable and explainable, you must also ensure that the monitoring and optimization of your model are scalable and efficient to avoid overtraining. One of the use cases of machine learning as a technology is to connect the dots of different data sources to gather intelligent insights – with the help of cloud computing, the ability to get used to new technologies more comfortably and connect other rising technologies. All of those insights I gained wouldn't have been possible without the thought of democratizing technologies.