Improve strategic decision-making and digital transaction experience, gain customer insights, and understand consumer purchasing behavior.
Fintech & Insurance
The financial sector is one of the largest users of digital technologies and a major driver in the digital transformation of the economy. Financial technology aims to both compete with and support the established financial industry in the delivery of financial services. Financial and insurance firms were the early adopters of the mainframe computer, relational databases, and have eagerly awaited the next level of computational power: AI.
Learn More
Use Cases
How we can leverage ML in Fintech & Insurance
Loan approval automation
Regulatory compliance
Investment prediction
Credit scoring
Personalized offers
Fraud detection
Insurance pricing
Claims reserve optimization
Frictionless Identity Verification
About the client
Paigo is a fintech company in the consumer lending space that caters to the unbanked. Through an ecosystem of brands, it offers 100% financial products for different client segments. Only in its first few years of operation, Paigo managed to capture +100.000 clients. Located in Latam, it is looking to expand across the region and was recently acquired by one of the largest banks in the world, the Brazilian Banco Itaú.
Challenge
Its core value proposition was the simplicity and fully digital experience in the lending process. They wanted this to be reflected in the loan application and payment collection processes. The solution focused on catering to the Argentinian market.
Results
Along with the client, we implemented a digital frictionless identity validation process using social media and Whatsapp, based on Computer Vision techniques. Identity was confirmed using photographs of the person and national ID documents and validated against trusted sources.
Moreover, a contact management system was implemented using Natural Language Processing, allowing our client to proactively engage with users through digital channels for payment collection purposes. It also has a dashboard that allows to schedule a given set of rules so as to send messages automatically and visualize all relevant contact metrics.
24
hrs
To turn around lending decisions
Impact
By leveraging AI, fintech companies can turn around lending decisions in less than 24 hours, compared to the long approval process that occurs in traditional financial institutions.
Impact
By leveraging AI, fintech companies can turn around lending decisions in less than 24 hours, compared to the long approval process that occurs in traditional financial institutions.
24
hrs
To turn around lending decisions
Technologies
Python, Pytorch, Transformers and visualization techniques.
Transformers
Visualization techniques
Credit Scoring and Churn Prediction
About the client
PedidosYa is part of Delivery Hero, a German multinational online food-delivery service. The company operates in 50+ countries internationally -spanning Europe, Asia, Latin America and the Middle East- and partners with 500,000+ restaurants. PedidosYa is the market leader for food delivery in Latin America. It employs more than 5,000 people and has a market valuation of over 3 billion dollars. Trusted by millions of users and thousands of partner stores, it has recently expanded to the fintech market.
Challenge
Our client was looking to set up a new business line to offer B2B loans to eligible partner stores. They identified an opportunity in a market where access to funds for micro, small, and medium-sized enterprises is limited and at high interest rates. This initiative would serve as an incentive for vendors to reinforce engagement with the platform and a key differentiator from other platforms. The main objective was to increase client loyalty and reduce churn.
Results
We partnered to create a credit scoring model and a churn prediction model to select eligible partners. The aim was to score each partner based on its probability of default and churn. A frictionless process was designed, so credit could be requested after going through a simple workflow and disbursed in less than 48 hours. It involved the development of a backend analytics tool for loan tracking and monitoring. For the purposes of payment collection, it was determined that installments would be collected through a direct debit instruction into the partner ́s bank account or debit card.
We designed and trained custom machine learning models for both the churn and credit scoring risk models. We identified relevant metrics to decide on the loan offerings and worked on the explicability of the features and models, looking for example at the Expected Maximum Profit and the utility matrix metrics among others.
2.5
Billion unbanked people globally
Impact
Credit scoring is being applied in multiple countries reducing churn and increasing loyalty and recurring revenue. With 2.5 billion unbanked people globally and less than half of the banked population deemed eligible for lending, there is a window of opportunity for AI-enabled credit-scoring solutions.
The MVP for a first country was built in three months and the first loans were made right after the MVP was completed. Expansion to new countries followed right after.
Impact
Credit scoring is being applied in multiple countries reducing churn and increasing loyalty and recurring revenue. With 2.5 billion unbanked people globally and less than half of the banked population deemed eligible for lending, there is a window of opportunity for AI-enabled credit-scoring solutions.
The MVP for a first country was built in three months and the first loans were made right after the MVP was completed. Expansion to new countries followed right after.
2.5
Billion unbanked people globally
Technologies
Python, Tensorflow, Keras, Google Vertex, Airflow, BigQuery
Identity Verification and Fraud Detection
About the client
With more than 1M clients and over 250k loans granted in Argentina, Perú and Uruguay, Prex offers a digital platform for financial products and services. It includes local and international purchases, transfers, service payments, loan applications and tools to manage personal finances. It aims to democratize access to financial products and services across LATAM, promoting financial inclusion and cross-border payments. It was recently acquired by one of the largest banks in the world, the Brazilian Banco Itaú. It competes in the same space as Nubank and Revolut and is a partner of multinational Mastercard.
Challenge
Prex had an engineering team that wanted to train, empower and expand with the support and experience of an external partner in machine learning. For that purpose we worked together, transferring knowledge and implementing their first AI projects together.
We worked with our client combining different modalities:
We worked with our client combining different modalities:
- Knowledge transfer & Workshops on ML project management, deployment, models and architectures
- Consulting and support from the Marvik team
- AI team leverages to speed up internal developments.
Results
One of the projects we collaborated on was the streamlining of the loan application process using Computer Vision Techniques. This involved the development of an identity validation system that used Optical Character Recognition (OCR) to extract text information from loan applicants’ IDs. This was validated against personal photographs based on a comparison tool that yielded the level of similarity. OCR was also used to develop a fraud detection system that involved automated face recognition and image processing to match the identity of the person.
Moreover, we created with them a solution using a natural language processing solution to help them automate part of their customer support. It had the added complexity of having a large diversity of use cases and limited training data, plus the fast-paced evolution of the business needs that the technology had to keep up with.
42
K
Victims of identity fraud
52
Billion in losses
Impact
Identity theft is deemed as a growing problem, as the number of fraudulent transactions and data breaches continues to rise. Almost 42 million Americans were victims of identity fraud in 2021, costing consumers a total of $52 billion in losses.
Additionally, AI-enabled chatbots can significantly help to reduce customer response times by up to 30% on average. This is especially relevant since repetitive queries amount to almost 80% of total queries.
Impact
Identity theft is deemed as a growing problem, as the number of fraudulent transactions and data breaches continues to rise. Almost 42 million Americans were victims of identity fraud in 2021, costing consumers a total of $52 billion in losses.
Additionally, AI-enabled chatbots can significantly help to reduce customer response times by up to 30% on average. This is especially relevant since repetitive queries amount to almost 80% of total queries.
42
K
Victims of identity fraud
52
Billion in losses
Technologies
Python, Golang, React, AWS, Azure, Google Cloud
Get in touch with one of our specialists.
Let's discover how we can help you
Training, developing and delivering machine learning models into production
Contact us
Got a project?
Let's talk
About us
Industries
Expertise
Services