AI & Data Science: The roles and skills that companies need

There has been a significant evolution in the level of sophistication in the use of analytics, machine learning (ML) and artificial intelligence (AI) in German companies in recent years. This is something we see every day in our customer projects. While a few years ago, proof of concepts and feasibility studies on the use of AI and analytics were still predominant, today machine learning and (generative) AI are often used operationally: from the automated processing of customer service inquiries to AI-based scrap control in the manufacturing industry and personalized healthcare.

Key roles in modern AI & data science teams

Data science and artificial intelligence are team sports. A broad spectrum of competencies and skills is required to successfully manage these often complex projects and to operate the necessary digital solutions in a stable and sustainable way. The teams should therefore include at least the following roles:

Role: The individual who assumes this role develops strategies and tactics to achieve the target of becoming a data-driven company. They identify where the company can gain an edge over its competitors by using and evaluating data in business processes. The Data Strategist’s job is to keep an eye on the bigger picture and bridge the gap between business and technology.

Competencies:

Strategic thinking and planning skills

Business know-how and knowledge of industry trends

Expertise in data analysis and interpretation

Communication and presentation skills

Stakeholder management and persuasiveness

Example: A retailer is looking to increase sales. The Data Strategist initiates a “Customer 360” project to analyze sales and customer data with the objective of identifying trends in purchasing behavior. From this, they then derive actions, such as how customer loyalty can be improved with personalized advertising campaigns.
Role: The DataOps Specialist is the builder, so to speak, who creates the technical foundations. These specialists are responsible for developing and implementing the data infrastructure, managing databases and providing data for AI solutions. They are the people who make the data work. This role also includes specialized profiles such as Data Architects or Data Engineers.

Competencies:

Database design and development

Data modeling and integration

Programming skills (e.g. SQL, Python)

Knowledge of cloud-based data tools such as Azure Data Factory or AWS Glue

Automation know-how (e.g. with the help of scripts or tools such as Apache Airflow)

Example: In the previous example project, the DataOps Specialist agrees with the Data Strategist and the modeling experts (see below) on what data is needed for what purpose. They advise on the provision and use of data and implement data pipelines to process data in real time.
Role: Here we have an expert in mathematical modeling, machine learning and (generative) AI. Their main task is to develop complex machine learning models that bring algorithms to life. This role comes closest to the “classic” data scientist with a focus on mathematical-statistical modeling.

Competencies:

Advanced knowledge of machine learning and deep learning

Expertise in specific ML areas such as NLP or generative AI/language models (LLMs)

Programming skills in Python and knowledge of ML libraries such as TensorFlow or PyTorch

Statistical analysis skills and mathematical understanding of algorithms

Creativity and problem-solving skills, especially when modeling complex problems

Example: In our example project, AI specialists are responsible for analyzing sales and customer data. They develop a recommendation system based on this, for example, which analyzes customers’ shopping behavior and makes tailored suggestions.
Role: This role ensures that everything runs smoothly from a technical perspective. An MLOps Engineer develops and manages the necessary machine learning and analytics infrastructure and ensures that the developed ML or AI components can be smoothly integrated into existing processes.

Competencies:

Understanding of analytics infrastructure and architecture

Experience with cloud computing platforms (e.g. AWS, Azure, GCP)

DevOps knowledge and experience with containerization technologies (e.g. Docker, Kubernetes)

Ability to implement MLOps practices and processes

Knowledge of machine learning and deep learning

Example: The MLOps Engineer ensures that the developed recommender system can be operated securely as a micro-service. They create a digital “living space” for the application.
Role: Last but not least, we have those who put the models into practice. These developers build the specific applications and systems that use machine learning. They ensure that the models are seamlessly integrated into the processes.

Competencies:

Knowledge and practical experience in the development of software, apps and systems

Proficiency in ML modeling and implementation

Knowledge of all programming languages used in the company

Knowledge of interfaces of the software used in the company, especially regarding the integration of ML services into the systems

Knowledge of databases and APIs for the integration of ML models in production environments

Example: The ML developers ensure that AI is ultimately deployed in the customer interface and that an enjoyable user experience is made possible.

How team members work together optimally in data & AI projects

Besides the skills and roles described above, effective collaboration between data scientists and AI experts also requires all team members to work together seamlessly. This begins with brainstorming and continues right through to the implementation and maintenance of AI and analytics solutions. A typical workflow might look like this:

1. Brainstorming and concept
During this stage, the data strategists gather information about the company’s business goals and challenges. They conduct market analyses where necessary and identify opportunities to optimize and improve processes and products using data from a business process perspective. Together with the AI Specialists and MLOps Engineers, they work to develop concepts for data-driven solutions that support the achievement of business goals. These solutions are then to be implemented in the processes.

2. Data acquisition and pre-processing
The DataOps Specialists lead the way in this area. They identify and obtain the data required for modeling and operating AI solutions from various sources. They also undertake the necessary data cleansing and pre-processing to ensure that the data is suitable for modeling. Data quality and data availability are key challenges in almost all data projects.

3. Model development and validation
This is where the AI Specialists come into play again. They develop and then train the machine learning models using the data provided. Meanwhile, the MLOps Engineers work on the development of validation strategies for the performance of the mathematical models, i.e. they ensure that the model performance is evaluated.

4. Implementation and deployment
The ML Developers now take the lead. They integrate the models that have been developed into production environments and develop applications and interfaces via which the models can be addressed and used to improve business processes. The MLOps Engineers support the implementation of monitoring and maintenance processes to ensure that the models remain stable during operation.

5. Operation and optimization
The workflow does not end with implementation. The team continuously works on monitoring model performance and identifying opportunities for improvement. The MLOps Engineers play a central role in identifying potential areas for improvement in terms of model performance and infrastructure.

Blog

How to make sure machine learning projects succeed

More and more companies are looking to optimize processes, products and services with machine learning. The desire to introduce and invest is there, but there […]

Strategy & Data Driven Business Artificial Intelligence +1

Success factors and best practices for data & AI projects

In practice, clear communication, close coordination and ongoing training are important factors for successful collaboration. The team needs to fully understand the business objectives and the requirements of the various stakeholders both inside and outside the company, as well as being able to react flexibly to changes and ideally always being one step ahead. How does a data team become and remain successful? These are our lessons learned from numerous projects:

Clear communication and transparent cooperation:
Open communication between all team members is essential for the success of a data science team. Each team member needs to understand how their role contributes to the overall outcome and be able to openly discuss ideas and challenges. Regular meetings and status updates help to keep everyone involved up to speed.

Flexibility and adaptability:
The world of data is undergoing constant change, be it due to new technologies, changing business requirements or external influences such as changes in legislation. Successful data & AI teams are able to adapt quickly to new circumstances and react flexibly to changes. As a rule, they pursue an agile approach to work.

Continuous training and development:
In order to keep abreast of constantly evolving technologies and best practices, it is crucial that team members continuously expand their knowledge and skills. Further training measures, training courses and participation in conferences and workshops are therefore crucial for the long-term success of the team.

Stakeholder involvement:
Data projects are rarely isolated endeavors, but often influence various departments and interest groups within the company. It is therefore vital to involve all relevant stakeholders in the process from the outset, to understand their requirements and to ensure that the solutions developed meet their needs.

Clear objectives and measurability:
Every AI & data science project should have clear goals and KPIs that can be used to measure success. By setting common goals from the outset and regularly reviewing progress, the team can identify potential obstacles at an early stage and take countermeasures in good time.

If you would like to find out more about how to set up the data & AI team in your company effectively or if you would like support in certain roles, please contact Dr. Markus Knappitsch and his colleagues: You can contact us here.