Summer of AI
2022

The 2022 edition of the series was the first to be held in person. Thus, we had the chance to organise much more exicting events for the participating firms and students, such as an Hackathon and Campus tours.

With 9 participating companies and 45 students being part of the edition, 4 award categories were part of program.

Students!Organisations!

Participating organisations

Our partners

Technology partner

Media partner

Our latest news

Shell

Introduction
Climate change necessitates a transition of our energy system towards renewable sources. Since the energy transition is too big for a single company to tackle alone, Shell has recently opened the doors of its Energy Transition Campus Amsterdam (ETCA) to start-ups, scale-ups, academia, and mature companies.Shell believes that digitalisation and AI are critical enablers to support our ambition to be a net zero energy company. We want to reduce our own CO2 emissions and help our customers reduce theirs. ETCA is a frontrunner and an example case of this effort. The campus’ electricity originates from on-site solar panels and Dutch wind farms. Lighting is fully LED-based. Heating is managed via underground hot/cold storage.This project focuses on various aspects of ETCA’s electricity consumption. Data science will play a crucial role and participants will have the opportunity to make a real and direct impact.

Research questions
• What kind of power storage (e.g. battery packs) do we need to shave off our peak demand from the grid (grid relief)?
• Can we give recommendations on energy efficiency measures, e.g. building insulation?
• Can we model our base load to serve as a reference for energy improvements?
‍
Project setup
Students joining Shell during the Summer of AI will be challenged to answer the research questions by modelling the electrical structure of ETCA and its main loads. Not all submeter stations are currently able to be logged remotely but more detailed load monitoring can be developed and installed. This project touches upon the various phases of a data science project and has two simultaneous intertwining pillars.

What we expect from you:
• Shell ETCA, Amsterdam (≥ 1 day / week) Virtual (rest of week). Align days within team.
• Commitment: ≥ 4 days per week
‍• DS/AI/technical students with experience or a strong interest in DS/AI. Permission to work in NL is required
‍• Internship compensation and travel expenses provided.‍

KPN

The Future of Marketing
‍
Data-driven decision making is becoming a more and more important topic in marketing to ensure better personalization for the customer. The ability to quickly experiment with various types of offers and communication channels, receive feedback from the customers, and identify the most rewarding promotion strategies is crucial to obtaining a competitive advantage. KPN has set the ambition to become more data-driven and increase model-based personalization in marketing. To achieve this, we are working with Reinforcement Learning, namely contextual Multi-Armed Bandit models (MAB). This brings challenges such as real-time decisioning implementation, continuous model improvement, and creating the right features from high-dimensional customer context data. In this case, you will get the chance to improve MAB models, to ensure that all customers receive the best offer at the right time, given their individual characteristics. We will discuss the challenges and problems that we face in implementing these solutions and give you chance to improve the quality of our Multi-armed Bandit models. Ready to rack your brain and think about the future of marketing? Join our case and let yourself be heard!

What we ask of you
• 4 days per week (non-weekend), at Amsterdam Office (Sloterdijk) 2 days a week.
• Preferred master student
• Compensation provided and travel expenses are reimbursed if you are not in possession of student OVcard.

Tata Steel

Introduction
Tata Steel is one of leading steel producers with steelmaking in the Netherlands, and manufacturing plants across Europe. The company supplies high-quality strip products to the most demanding markets, including construction and infrastructure, automotive, packaging and engineering. Surface quality is an important aspect of product quality, where we strive for zero defects. Our installations provide a whole range of possible causes of surface defects: scratches, smears, stains, pits of all sorts and sizes.
Crucial in achieving zero defects is a large number of automatic camera inspection systems which we have installed in several places in the process. The system captures every surface defect and classifies the defects, which are used to determine whether the strip needs to be reworked. By correlating the defects to process data of our installations we can identify the root-causes.
The current defects classification model development has been supervised, i.e., showing the algorithm what a scratch looks like and then refining what kind of scratch it is. The current model classifies the defects based on features derived from the defects image. The ideal classification model would enable us to prevent defects from reaching customers automatically.

Challenge
Our challenge to you: can you create an defects classification system that detects different types of defects that outperforms our current system.

Your summer of AI at Tata Steel
During the summer, you’ll have the ability to create a computer vision model that detects and classifies defects unsupervised. You will have access to camera data of one of our finishing lines and will be supported by surface experts to validate the findings. At the end of the summer, you have learned how to create a proof of concept and work with large scale data systems, such that at the end of the summer, we can start implementing this system.

What we ask of you
• 4 days per week (non-weekend) of which min. 1 day per week on our site at the Tata Steel IJmuiden.
• Preferred master student in STEM related field with experience with ML/AI.
• Compensation provided and travel expenses are reimbursed if you are not in possession of student OVcard.‍

ABN AMRO

Introduction
‍“Banking for better, for generations to come” is the purpose of ABN AMRO Bank. We help to shape the future by taking our role in society seriously and working together with our clients to tackle the challenges of our time. We believe that all our clients will be affected by the transition to sustainability in one way or another in the years ahead.
‍
‍Challenge
‍One of society’s initial steps in this transition is creating a sustainable energy supply. Therefore, our challenge to you is to use AI to identify and size opportunities for solar panels for retail and SMEs in the Netherlands. The challenge is twofold. From a business perspective, we need to scope the problem and detect opportunities. From a data perspective, we’re dealing with variety of components such as data gathering, processing, model creation and validation, that will proof that the solution is feasible and within reach. The resulting model can be directly applied to a variety of use cases and will help us to serve our clients better in their journey towards a sustainable future.

‍Your summer of AI at ABN AMRO
‍You’ll join the AI team of the Strategy and Innovation Department and experience the life as a data scientist at ABN AMRO. You and the team will be mentored by our data scientists that guide you through the phases of the project, ensure you to work agile and deliver something to be proud of. At the end of your summer, you have learned how to tackle a problem head on and deliver a working proof-of-concept to internal stakeholders.

‍What we ask of you
• 4 days per week (non-weekend) of which min. 1 day per week in our office at the Zuidas Amsterdam (if government policies allow).
• Preferred master student in STEM related field with some experience with programming (Python) and ML/AI.
• Provide a VOG at the start of the program (https://www.justis.nl/en/products/certificate-of-conduct)
• Internship compensation and travel expenses provided.

Company.info

Introduction
Company.info is part of the FD Mediagroep and market leader in Dutch business information and insights. We provide our clients with qualitative, up-to-date and in-depth business insights and the latest company news. We provide solutions for every business professional, who wants to build strong business relationships and increase their sales. We make our data available directly via an online portal, integrated into systems via APIs, as a basis for analysis in online dashboards and as a data delivery. This way we help professionals to be more successful in doing their business.

‍Challenge
At Company.info we want to know if companies are mentioned in the news. We solve this by using entity recognition and entity linking. We currently have a good solution, however there are some specific situations in which we would like to improve. Our current setup is language dependent, it works well in Dutch, but not so much in German. Also we've noticed that the Dutch system doesn't always work well in formal / legal texts. We want the students to investigate how we can improve in these cases. This could be done by making the current model more flexible, but we can also train models specifically for the other cases.
A Company.info we want to know if companies are mentioned in the news. We solve this by using entity recognition and entity linking. We currently have a good solution, however there are some specific situations in which we would like to improve. Our current setup is language dependent, it works well in Dutch, but not so much in German. Also we've noticed that the Dutch system doesn't always work well in formal / legal texts. We want the students to investigate how we can improve in these cases. This could be done by making the current model more flexible, but we can also train models specifically for the other cases.
‍
‍Your summer of AI at Company.info You will work closely with our data science team and will have a dedicated supervisor. You will also be given the opportunity to share you work with other people in our organisation during one of our demo sessions. Here you can get feedback on the practical relevance of your solutions. The end goal of the project will be to have a model that is ready to be integrated in our production systems.

‍What we ask of you
• 4 days of a week, we expect you to be primarily at our office in Amsterdam
• Preferred master students in AI, Data Science, Computer Science or another relevant field
• Travel expenses will be compensated

South Pole

Introduction
South Pole is an energetic, global company offering comprehensive sustainability solutions and services. With offices spanning all continents across the globe, we strive to create a sustainable society and economy that positively impacts our climate, ecosystems and developing communities. With our solutions, we inspire and enable our customers to create value from sustainability-related activities.Our Digital & Data Science team helps our clients and South Pole design and execute the most ambitious and scalable climate impact initiatives whilst reducing the administrative effort to zero. Our unique track record of running projects across the globe and delivering verified, high quality carbon credits provides us with valuable data we can leverage to identify new, high impact initiatives. We help companies on a day-to-day basis to drive their carbon neutral strategies with their customers, suppliers and their employees through analytics and sustainability software solutions. If you have the skills, ambition and drive to use the best of digital and AI to protect ecosystems, restore forests, drive sustainable agriculture, and enable climate zero journeys across the globe then consider being part of our Summer of AI.
‍
‍Challenge
To improve the way we monitor the impact of our AFOLU (agriculture, forestry and land use) projects, we are currently exploring a variety of ML applications on remote sensing data:
• Can we estimate the amount of carbon sequestered in a forest purely based on remotely measured data?
• Can we synthetically increase the resolution of historic satellite imagery to improve the estimation of the carbon baseline prior to our intervention?
• Can we automatically identify areas that show structural loss in carbon sequestration that would greatly benefit from an intervention?
• Can we fuse optical and radar satellite imagery for improving the reliability of digital monitoring solutions for forestry projects?
‍‍
We are currently only scratching the potential of each of these applications and can use all the help we can get to further explore their potential. Depending on your passions and strengths we would look at which of these challenges would be the best fit for you.

What we have in store for you
‍The opportunity to contribute your machine learning skills to real world projects addressing climate changeThe opportunity to learn from and work with colleagues all across the world working on climate change, not just through AFOLU projects, but also through biodiversity, e-mobility, carbon capture, plastics and other types of projects. A stipend of 500 EUR.

What we ask of you
• Availability of at least four days a week with the ability to join us in the office in Amsterdam at least 2 days per week
• In an MSc degree of having finished an MSc degree with at least some experience applying AI/ML to remote sensing data

Aegon

Introduction
In the past few years Aegon has invested heavily in the data science capabilities of the company, which has led to a central Analytics & Data Science department with approximately 40 data scientists. There is a variety of backgrounds in the department with academic team members who are econometricians, mathematicians and physicists. The Analytics & Data Science department consists of 4 teams with a different focus: commercial pricing, process and journey analytics, customer research and risk analytics. The commercial pricing team has access to data like competitor price information, quote and sales data and product characteristics. As customers and distribution channels react differently on value and price characteristics lots of optimization possibilities arise.

Challenge
‍As the pension market in the Netherlands is rapidly changing from defined benefit to defined contribution pension schemes, the individual annuity (DIP, direct ingaand pensioen) product is an increasingly important product for Aegon and its customers. An annuity is the product that disburses monthly when the lump-sum payment from the defined contribution scheme at the retirement age is made. Within the DIP pricing team many efforts are made to build an infrastructure that accelerates and automates the process that calculates and lists the price of the annuities. This price is currently based on the actuarial cost price estimate and price position in the market, no other factors such as distribution channel or customer characteristics are included yet. As the market is growing, it is increasingly important to offer an attractive price that contributes to profitable growth of sales for Aegon. To do so, we first need to have a price elasticity model that tells us which characteristics drive a customer to buy a DIP product.

‍Your summer of AI at AEGON
‍During the summer, you have the possibility to create a new state-of-the-art AI model that predicts the conversion of the DIP product. You will have access to a database with competitive price information, an overview of the products offered at competitors and access to our datalabs where the data is available and can be used within Python or R. At the end of the summer, you have learned how to create a proof of concept and bring your model to production, such that at the end of the summer, your model is ready to go into the field.

‍What we ask of you
• 4-5 days per week of which at least 2-3 days in our office at Aegonplein The Hague together with the team
• Internship compensation is provided

Nederlandse Spoorwegen

Introduction
On an average day, NS facilitates 1 million journeys per day. Crowdedness of trains was always a topic for customers but during Covid, we have seen a change in the experience of crowdedness. With fewer passengers, it felt more crowded. We also see that our predictions need to react more to what is happening in the world. During Covid, press conferences of the government had a direct impact on the number of passengers. To cope with this new situation, we have developed a new deep learning model to predict the crowdedness of trains. A new crowdedness prediction is made every night for the coming three days and during the day an update is made for the rest of the day. These predictions are a combination of different models as we have learned that there is not a single model that performs best for all circumstances.

Challenge
One of the biggest challenges with our current models is events like soccer matches and festivals. What is the impact of these events on the number of passengers? We need your help to predict the crowdedness of trains when an event is taking place. When are people traveling to the event? When are they traveling back home? A football match and a concert have a fixed end time, and a fair (like Huishoudbeurs) has a more distributed travel pattern. Visitors come and go as they wish. Can you crack this case for each? Each event seems different with respect to the number of visitors and the travel plans. We need to find a way to add these visitors of events to our prediction model. When, for example, the national soccer team plays in the Amsterdam Arena, the impact is nationwide. We see a major impact on the trains from Amsterdam and Utrecht, but the ripple effect is noticeable everywhere. Especially after the match when everybody wants to go home. At the same time.

Your summer of AI
During the summer you will work with experts in the field of crowdedness predictions on predicting the impact of events. You will have access to relevant data and models that are currently used for crowdedness predictions. We also provide a scalable AI platform in the cloud for exploration and analysis, from statistics to advanced deep learning models. Four days in a week you will work at our headquarters in Utrecht at this problem in close cooperation with our experts of which two field experts will accompany you on a daily basis.
‍
The assignment:
• Commitment to working 4 days a week on this challenge4 to 6 master students with Python and AI experience
• The work location is the NS office in Utrecht (next to Utrecht Central station)
• Internship compensation is provided
• A VOG is required to start

DSM

Uncovering drivers for fish oil purchase recommendations
One of the DSM businesses based in Latin America relies on the availability of fish derived oil. This particular oil is available mostly during the fishing season. The price of the oil changes depending on how well the season is going. If there is scarcity of fishes, the price may go up. If the weather is mild, making the fishing operations easier, more fishes are caught, which may make the price go down.

The business using the fish oil needs to take decisions about the best time to buy it, and in which quantities, based on the forecasting about the fishing season. A number of structured and especially unstructured data sources provide information which can be utilized to make and improve the prediction, but the relation between these data points and the price trend is not yet established. There is also the possibility to enrich the data set with new sources, which in turns should improve the accuracy of the forecast (wait and buy later because the price will decrease or buy now because the price will increase).

The team should create a machine learning model which, taking the available information (structured & unstructured) as input, plus additional information such as weather forecasts, etc. that produces a recommendation about the purchase of the fish oil.

What we ask of you
• 2-3 days per week commitment.
• Travel reimbursement to DSM offices (location can be decided based on student preference)
• Internship compensation provided.‍

Summer of AI2022