Creation of quick prototypes, verification of product, technological and business hypotheses and implementation of infrastructure projects
Integrating into the bank's business processes and adapting our technologies to the specifics of the customer's processes.

FEATURES OF WORK WITH CLIENTS IN THE FIELD OF FINANCE
Personal Information
We have to work with personal data - either we anonymize or work on the bank's side.
Data Engineering
You have to work a lot with data and build them into single marts around which subsequent modelling will be designed.
Bank side work
Access to many systems and integration into the circuit is only possible on the customer's side.
Multidisciplinary team
You need to work with both models and integrations, with big data, with process optimization - this requires different specialists.
Complex integrations
A bank is the merger of various departments, each may have its own specifics, you have to integrate with different databases and data sources.
Dealing with uncertainty
Business requirements are often poorly formalized, as it is necessary to achieve performance indicators, but there is no clear road there.
WHAT DIRECTIONS WE DEVELOP FOR BANKS
WE HAVE TWO MAIN COOPERATION DIRECTIONS
ML product prototyping from scratch
We know how to...
  • start the product "cold"
  • make a tech MVP without data
  • collect this data
  • come up with a small prototype that already delivers value
Because ...
  • we have experience working with tasks at the "zero starts"
  • we understand how to quickly reduce a task to a solvable one
  • we know what is the difference between the technologies used in standard machine learning with the presence of a large training sample from the zero stage of product development, when there is no data at all
  • we use both expert knowledge and specific technological solutions
  • we are able to collect and annotate data using crowd-sourced platforms and specialized companies
Interesting to create a product
Infrastructure setup, ML-Ops
  1. Building сontinuous models integration
  2. Tracking the quality of models
  3. Debugging a continuous stream of data
  4. Setting up prototyping, testing and production outline
We have already built the infrastructure within our team and in several R&D divisions of large companies. We use up-to-date ML-Ops frameworks - from Airflow to Wandb, we are able to integrate solutions with Spark / Hadoop. In general, we are doing everything to establish transparency and provide clear tools for teams within companies.
Interesting to set up the infrastructure
IN COOPERATION WITH OUR TEAM
Up to 5 million rubles
The bank saves on gathering a team to start a new direction
Up to 3 months
Developing prototypes for testing the hypotheses reduces the time
Up to 3 million rubles
Cost savings in the case of stuff reduction
Up to 6 months
The development and implementation of the required infrastructure is faster
EXAMPLES OF THE PROJECTS FOR BANK PARTNERS
WHAT TOOLS WE USE TO CODE
What we use to code
Main programming language: Python
Working with data: PostgreSQL, MongoDB
Distributed computing: Spark, Hadoop
Notebooks: Colab, Jupyter, H2O
Model packaging and data flow control: Kubernetes, Docker, Airflow.
Profile frameworks
Neural network frameworks: PyTorch (prototyping), TensorFlow (production), TensorFlow lite (import to devices)
Working with data: pandas, NumPy
Profile libraries: BigARTM, TopicNet, gensim, nltk, DeepPavlov, SpaCy, OpenCV, SciPy, etc.
Experiment management: Wandb, MLFlow, Tensorboard.
Code Style and solution delivery
  • Styling your code as python scripts using PEP 8
  • High level of code readability

Delivery options:
  • Docker + REST API;
  • Web-service + Frontend (for example, Python Flask + React);
  • Python-scripts;
  • Python libraries.
Order a project
To start the project, we need to talk. Just fill in the fields below and we will contact you.
Promo code
Order a project in one click
To get started, we need to talk. Just tell us about the project and specify your contacts. We will contact you to discuss everything.