We have published a video and demo of our RAG solution. It significantly outperforms the traditional RAG approach and allows creating more intellectual AI assistants. You can try our AI assistent right now!
The implementation of the customer reference data management system (CDI, MDM) in Eurasian Bank JSC has been completed. The generation of reference data objects for the customers, documents, and addresses based on data from five automated systems of the Bank has been implemented. Data processing occurs in a near-real time mode and includes extraction and transformation, cleaning, validation, and consolidation of information from different sources. Reference data can be used by business applications using the platform API or by synchronization (replication) using subscription. The total volume of data processed by the system is more than 100 million records. The flexibility of setting up information processing rules in the EKG platform made it possible to quickly implement changing functional requirements as the project progresses. The advanced capabilities of the platform for consolidating information made it possible to generate reference records using only the most relevant and correct data for each customer. The result of the project was the creation of a complete and validated array of data on the Bank's customers, which can be used in the execution of operational business processes and in analytics, as well as the emergence of tools for the continuous support of the quality and completeness of reference data.
We have completed our work on creating an address recognition service for JSC Kazpost. The service cleans, standardizes and validates addresses of international postal items
We have presented our product at the Data Summit 2030 MEA in Dubai. A lot of interesting conversations with the possible customers and partners. Looking forward to continue promotion on the Middle Eastern market.
The fall conference season continues with Go Digital Eurasia. We held dozens of meetings with CIOs and CDOs interested in data management solutions
Our stand has attracted the attention of CDOs and specialists of the largest Kazakhstan companies at the main data management conference of the year
The project on analyzing the text of construction industry normative documents using NLU was implemented for KazRICA JSC
We have published a video about the AutoML tool in the EKG Platform that enables businesses to utilize machine learning tools.
EKG Platform integrates with OpenMetadata to implement a data catalog and automate ETL rule creation.
A demo version of the Kazakhstan address cleaning service has been published. The service splits the address into components and determines the KATO code.
We have published a video covering the features of EKG Language Processing, a NLU tool that can be used to handle tech support incidents
A pilot project on automatic classification of incidents received by the technical support service was implemented. Among tens of thousands of text descriptions of incidents, groups of messages close in meaning were identified. An algorithm that extracts the exact meaning of statements from the text was applied: first, the ML model is used to recognize the grammatical structure of the phrase, then the conceptual structure of the statement is formalized using the SKOS conceptual model. In this way, similar incidents can be combined even if their essence is described by different words. The model requires no prior training and no manual creation of a typical problem classifier. The use of this tool will speed up the dispatching and processing of messages by the support service, timely detect and respond to mass violations of customer service, and analyze the quality of the organization's products and services.
A pilot project for calculating the optimal route of mail delivery for a logistics operator was carried out. Clustering methods were applied to distribute items between couriers, and an algorithm was implemented to calculate the optimal route for each courier, taking into account movement on the road network.
Release 1.22: the subqueries and aggregation functions, asynchronous validation and consolidation of data objects are implemented
A pilot project to create a single repository of reference information on bank customers using DataVera EKG Provider has been completed. As a result of consolidation according to customizable rules, a set of reference data is formed, which are distributed back to the source systems and provided to other consumers. The volume of processed data is millions of records.
Release 1.20: implemented asynchronous validation and consolidation of data objects
We have performed a pilot project for First Credit Bureau to consolidate and quality control data from multiple sources.
Release 1.17: implemented automatic generation of reference objects according to the rules specified in the model, taking into account data quality metrics.
At the Go Digital Eurasia conference in Astana, we presented our report "From data to knowledge: a holistic view of corporate IT in an era of accelerating innovation"
A pilot project on predicting the outflow of bank depositors using machine learning (ML) methods was carried out. A model was built to predict the probability of deposit termination based on behavioral and socio-demographic factors. Using the results of the model for timely contact with customers will allow a financial organization to retain a significant amount of funds on deposits.
DataVera became a participant of Astana Hub.
DataVera is a partner of the Profit Finance Day conference held in Almaty
EKG Platform 1.16 release: implemented value calculation and SPARQL functions on EKG side. Data quality indicator in EKG Explorer.
The "Data Consolidation on DataVera EKG Platform" video has been released;