Katharina P.

From Studying Statistics to Actually Applying it

After finishing my PhD in Statistics, I started my Data Science career at TUI in 2018. I joined a then newly established Data Science team that was created to implement Machine Learning projects in different business areas of TUI. The colleagues had just successfully setup their first projects in image and text recognition. With 2 Data Scientists and 1 Data Engineer the team was quite small but very motivated in pushing the application of Machine Learning models at TUI forward. We tried out different tools for the development and deployment of our Data Science powered features and products until we were confident with our choice. Furthermore, we had to develop common ways of working and guidelines on how to decide which projects to support and how to prioritize them. Nonetheless, we continue to challenge our infrastructure and ways of working constantly and adapt them if necessary to keep up with our needs that may change over time. To summarize, I would say that I have started my Data Science position at TUI as a statistician with only little practical experience. Over the years my skills in programming and dealing with the infrastructure have improved and I have gathered a lot of experience. This now enables me to guide and advice Junior Data Scientists who are new at TUI.

I Build Smart Data Driven Tools to Support the Business

I usually join different project teams to build smart data driven tools that support the daily work of my business colleagues. My business colleagues are the people who work directly with our TUI products and everything related to them, for example pricing, customer satisfaction or the content of our TUI website. As a Data Scientist, I work in an agile way and in a continuous feedback loop with the business colleagues. That means instead of implementing the whole solution and asking the users about their opinion at the end of the project, I set up and refine individual parts and features step by step and get their feedback after each step. This usually increases the acceptance of the products that I deliver and helps me to react on unexpected behavior of the model very quickly. I am typically participating in all steps of the development process of a project: from the analysis of the actual problem over the selection of relevant data and a suitable model type, the feature engineering and model training to the testing and deployment of the implemented solution. However, I do not have to work alone on all of these tasks since I am usually supported by business analysts that have a good understanding of the relevant data, database developers that set up stable data delivery pipelines to use the solution in production as well as data and platform engineers that support the deployment of the product.

The Work is Diversified and Challenging

I do not work in one specific business area but join different teams for some months to support them in setting up a data driven solution for a specific problem. It is very interesting to dive into a different business area every few months for a new project. As the requirements of the individual areas typically vary widely so do the solutions they need. I often have to teach myself new methods or dig deeper into the understanding of methods to check whether they might be suitable for my current use case. I really enjoy that my work is not only setting up the same model for each problem and that I often need to come back to my “classical” statistics skills to investigate whether my product is working as expected – and if not why. To keep up with the changes and progress in the machine learning world, I am always motivated by my manager to spend time on learning more about topics that might help me with my job. This might be Udemy or Coursera online courses to learn more about new methodological approaches and the theory behind them or courses to improve my skills to work with the common tool set. For example I spend some time to improve my programming skills in terms of making more use of object oriented functionalities or design patterns.

Bring Good Presentation Skills and the Will to Grow

I think you should be aware that communication and presentation skills, motivation and the will to grow and further develop your abilities are not only buzzwords but skills that are often underestimated but quite relevant for the work of a Data Scientist. I often have to explain and present the approaches I used and the results I obtained on various levels of granularity to different audiences. For example, when discussing with other data scientists in the company I need to be able to explain how a method works in detail. In contrast to that I have to focus on the business implications of my modeling choices when talking to the business colleagues or to managers. For the latter audience it is usually more relevant to understand the general idea of an approach rather than the details of how it is working. It is not easy to switch between these settings and to find a proper way of communicating to each of these different kinds of audiences but keep in mind that this will be a relevant part of your work as a Data Scientist. Additionally, lifelong learning is not only a buzzword at TUI. You should be motivated to actually stay up to date with the current state of the art Machine Learning procedures and bring motivation and initiative to continue your learning process. As a team, we are deeply involved in decisions that concern the way we work and the tools we use. Therefore, we have to keep track of new developments in the world of Data Science and Machine Learning and need to be willing to adopt new methods or tools if we expect them to be useful for our daily work.

Bringing Solutions into Production and Seeing People Work With Them

I am really happy that one of the major objectives in my Data Science team is to bring solutions into production and it makes me proud to see products I built with the project team being actually used by the business colleagues as part of their daily work. On the personal side I am very proud of having adapted successfully to rapidly changing business requirements and the technical landscape in terms of tooling and ways of working. After doing research at the university it was quite hard for me to switch the setting and to cope with the fact that working in practice requires a reordering of priorities. That is I had to accept that a good solution does not have to be the best solution in theory but the one that yields useful results while being stable and reliable in a productive environment. Furthermore, with the help of my colleagues I have managed to build up a community for Data Scientists all over the company in the last years. This allows people from different areas to get into contact and exchange about what kind of projects they are working on. We organize regular events like workshops on code improvements or the use of common infrastructure and deep dive sessions on the methods used in specific projects. We also organized a Hackathon in the last year to collect ideas for the further development of on an actual Data Science project. All in all, I can say that I have developed and improved many of my personal competences since having started as a Data Scientist at TUI: I have improved my knowledge about tools and infrastructure (especially on working with cloud resources), refined my programing skills and enhanced my organizational skills when it comes to collaboration with my direct colleagues in the Data Science team but also with colleagues in specific project teams.