We’re here to present our proposal for Challenge 1, which focuses on predicting default risk based on credit ratings.

Our approach explores and compares several algorithms found in recent research papers, aiming to identify the most effective model and understand how it can be improved.

By developing a reliable algorithm to evaluate potential clients, we can optimize client selection and reduce the risk of funding unreliable applicants.

This is crucial, because making financial decisions without accurate risk assessment can lead to significant economic losses. Our solution helps ensure that companies make data-driven, trustworthy decisions when granting credit.

We are Group 53, and we have developed a tool to help financial companies make better decisions.

Our solution relies on public information and previous operations to ensure the information and the reasoning behind our AI model is available to you at any time. We also offer the possibility to guide the software, giving you full control of the important financial decissions made by our software.

To sum up, we offer your company an easy-to-use credit scoring solution that does not compromise on transparency and user adaptability.

To overcome this challenge, we have developed an explainable AI credit scoring system that integrates both internal and external data to accurately estimate the probability of default.

Unlike traditional black-box models, our system provides transparent and interpretable results, expressed through a clear AAA–D rating scale. It not only predicts credit risk but also explains why, by revealing the key factors behind each score. This makes every decision traceable, reliable, and easier to trust.

We are a team of four engineering students with strong expertise in coding, finance, and artificial intelligence — united by the goal of making credit assessment more transparent, fair, and data-driven.

We all struggle with data, especially when it comes to analysing huge amounts of it.

Imagine having to deal with this problem every day. That is what most businesses need to do.

To help them (and each other too), we present an innovative way of classifying information into segments.

This will make it easier to find regularities, temporal trends or outliers in client databases.

We focus our work on companies interested in data analysis with an economic goal, as we are a team of computer science students passionate about turning data into strategic knowledge.

We’re here to present our proposal for Challenge 1, which focuses on predicting default risk based on credit ratings.

Our approach explores and compares several algorithms described in recent research, with the goal of selecting the most accurate model and improving it through our own optimizations.

Unlike many existing solutions that rely on a single predefined method, our system combines multiple approaches and adapts dynamically to different types of clients and credit profiles. This allows us to achieve higher prediction accuracy and better generalization across real-world data.

By developing a reliable and adaptable algorithm to evaluate potential clients, we help companies optimize client selection and avoid funding unreliable applicants.

This is essential, since making credit decisions without robust risk assessment can lead to major financial losses. Our solution offers a smarter, data-driven way to choose the right clients with confidence.

This risk is not only about the business’s balance; it's about analyzing the company’s cash flow, such as imports and exports, market flow, or the credit ratings in a country. Some businesses decide to ignore this problem and just lend money to "random businesses," and most of them end up in a very bad financial situation.

We are Group 27, and we have studied this problem by analyzing this data and working with AI. We present our solution: Our financial AI is trained with diverse datasets from emerging markets to identify not only high-risk borrowers, but also strong, reliable businesses that represent good investment opportunities

Our system combines large-scale financial and non-financial data to deliver accurate credit predictions with a very small margin of error. This makes it easier for lenders to understand how credit risk truly works and decide where their money will create the greatest impact.

In conclusion with our AI you are able to understand how this credit risk works and where to lend your money.

Thank you!

Imagine applying for a loan and only getting a number back — no explanation, no clue what it means, or how to change it.

That’s how most credit scoring systems work today: opaque, static, and hard to trust.

We’re building a fair and transparent credit scoring model that rates users from AAA to D, but actually shows how the score was made and how it can improve.

Unlike most models, ours aims to adapt in real time to macroeconomic shocks — things like inflation, interest rate hikes, or sudden market shifts — so the system stays fair even when the economy changes.

We also want our model to be understandable across languages and to draw on relevant external data, such as financial news or updated risk reports, to make smarter and more timely decisions.

In short, we want to make credit scoring transparent, adaptable, and human — turning “computer says no” into “here’s why, and here’s what you can do.”

Our project tackles Challenge 2: Automated Lead Discovery.

Today, banks and financial institutions struggle to identify small and medium enterprises (SMEs) that have foreign exchange (FX) exposure and potential for new business. This process is often manual, slow, and incomplete.

Our solution automates this discovery using natural language processing (NLP) applied to public data sources. It detects SMEs likely to operate internationally, then ranks them by their revenue potential, helping financial teams prioritize the most promising leads.

Think of it like a teacher spotting which students are ready for a debate, our system recognizes which companies are ready for FX solutions.

By turning unstructured data into insights, we empower institutions to uncover hidden opportunities, boost efficiency, and stay ahead in a competitive market.

We look at our client's behaviour using revenue, transactions, orders and products. From there, we create clear, human segments with simple profiles and revenue per client. Like a great teacher who adapts to each learning style, we adjust the offer, the timing, and the channel so every message feels relevant, not random. We use simple shopping signals like last purchase, how often they buy, how much they spend, and their usual product mix, and we apply K means, hierarchical clustering, and DBSCAN, reduce noise with PCA, and check quality with the silhouette score. The result is better targeting, less waste, and higher sales and customer value over time.

We are team 26 and we have developed and studied how you can maximize your efficiency by using cutting-edge technology like Machine Learning Algorithms and over 1 million data.

We distinguish ourselve by the use of scalable technologies and refined algorithms rather than the more traditional approach of using simpler algorithms that are now redeemed unusable because they cannot manage huge datasets.

Thank to our services, our clients have experienced an improvement of around a 30% increase in benefits. Due to client clustering and segmentation we were able to identify focus points and potential markets and clients.

In conclusion, we are a compromised team focused on creating long lasting relationships with our clients by using top-notch and trustworthy technologies.

A prime example of this is Ebury, a company that serves a diverse range of clients, from small businesses to large corporations with complex currency needs. As their client base grew rapidly, their one-size-fits-all approach became less effective. To address this, we propose a solution using data-driven segmentation, applying machine learning techniques to analyze transaction behaviors, client size, risk appetite, and growth potential. This approach will help businesses like Ebury better understand their clients, optimize resources, and create tailored solutions that enhance client satisfaction and drive profitability.

We are Group 31, and our solution to the main challenge is to create an automated lead discovery tool that uses public data and various natural language processing techniques to identify SMEs with high potential FX exposure. By analyzing data from company websites, directories and news articles, we can detect keywords and diverse patterns related to international trade, such as imports, exports, or other cross-border operations. Later on, we apply a model based on grading and scoring each company's FX revenue potential to help financial enterprises prioritize leads and improve the efficiency of their sales outreach. 

This way, we offer a solution that delivers the best results by identifying the most interested clients based on their key interests achieving greater accuracy than traditional methods. We’re looking for partners to collaborate on our automated lead discovery tool and validate its ability to identify and rank SMEs with foreign exchange exposure using real market data.  

However, it is an impressive first elevator pitch. Congrats! :)

Unclear solution: “We made a system” doesn’t say anything. In one short sentence, say what it does and how. Example:

They are very short specific

•  they mention that they have a lot of background- why should I trust them?

They don’t comment on their solution, just say that with their algorithm we can solve our problems, but why or how would I solve them?

Right now, getting a credit score today can feel confusing and unfair.

You’re told “yes” or “no,” but never why.

Our project fixes that. We’re designing a credit scoring system that rates people from AAA to D and clearly explains why they got that rating and how they can improve it.

Unlike most models, ours can adapt to real-world changes — things like inflation, market shifts, or new regulations. It also works across languages and can use news and external data to stay accurate and fair.

No more mystery scores or unfair surprises — just clear, adaptable, and trustworthy credit decisions that put people back in control.