Developing credit and operational risk management solutions for global financial
focus on processing large data arrays, introducing machine learning mechanisms and artificial
We use a narrow AI that allows risk officers to build, test and validate advanced credit scoring models.
IDF Lab has enhanced the credit scoring capabilities of its corporate
the integration of newly developed AI-based fraud-scoring engine.
The fraud-scoring engine works by analyzing information relating to
applicant’s authenticity and detecting atypical behaviour or anomalies in
The engine takes into account a wide range of factors including the time
spent filling in the form as well as the applicant’s web browser
The Application Scoring engine compares the data provided by the
filling out the loan application with data in other available databases.
The system identifies anomalies in the client’s application by
numerous variables, including the time taken and the precise mechanics of
the application fields.
The Application Scoring system has been developed using a combined stack of
learning algorithms, including Boosting, Random Forest, SVM, Logistic
The system is able to retrain scoring models instantly using mini-batch
optimisation. This prevents overloading of models when new data
IDF Lab has introduced a ML-based Customer retention system that
automatically grants each customer an individual loan discount by analysing
The propensity scoring assigns the borrower a special score that evaluates
to return to the service and, based on this, grants the customer a
The system analyses credit history data, personal traction of the
cooperation with the company, telecom and search engines data,
indexes, the customer's behaviour on the web site, and so
Collection Scoring is a system that is used when servicing
The system evaluates the probability of receiving payments from each
by analysing big data sets. It segments the borrowers into categories and
the most effective debt recovery strategies.
The system calculates the efficiency of each activity and test
the debt collection department.
Collection Scoring also allows risk officers to automate the information
exchange within the company and with external parties when working
product approval rate by eliminating fraudulent applicants
Currently, general-purpose computer vision models are most often used for document
The IDF Lab module is based on an ensemble of several neural networks,
which specializes in one type of object (passport, driver's license, bankcard,
allow processing documents much more accurately.
Financial planning and analysis
Financial planning and analysis (FP&A) is a cornerstone of the decision
process in the company.
We view our business as a mechanism which can be optimized while its constituent
such as marketing and credit risk management can be accurately
We spend a lot of time calibrating this model and then use it as a blueprint for
decision making. This includes pricing, risk management, marketing policy,
allocation and capacity planning. This is our nexus bridging the
with strategy and informing execution.
Client service automation
IDF Lab has developed and introduced a self-learning chatbot, which cut
services workload by a third.
The chatbot interacts with new customers and registered users and helps
the information required to determine financial product eligibility. It also
recommendations for the individual’s requirements and financial prudence.
Information is processed based on statistical matches covering a wide range
frequently asked questions. Thanks to the machine learning technology,
questions the chatbot is able to answer increases by several percent daily.
The chatbot works within the NLP
Processing) and NLU (Natural Language Understanding) AI frameworks.
Shape the future of finance. Take a look at our open positions and come join the fun.