Roman Khaykin

Valmarc Brand Protect is an application that can differentiate an authentic brand product from a counterfeit product. It comes to a verdict based on a large volume of data received from the manufacturer of the brand product.

As it is a data-centered application, it requires complicated database operations. Keeping this in mind, I designed the initial data model for the Brand Protect application and implemented it on two different platforms: MySQL and Oracle.

I designed the data load processes and implemented them using SQL*Loader, Shell scripts, and Python scripts. That included such amazing technics as parallelization and asynchronous processing to allow timely processing of complicated aggregations. I developed stored procedures and packages; some of the procedures, despite containing complicated logic, required a short response time—under 50 milliseconds. I also did a lot of SQL and database tuning to satisfy the requirements.

R4 Technology utilized a complicated mathematical algorithm to perform data analysis. A huge volume of data was processed using Oracle and PostgreSQL relational databases and NoSQL databases (HBase, Hive, and Impala).

I developed Oracle stored procedures and packages which were used by data analysis software to store and retrieve data from the database as well as implemented parts of the analytical algorithms. I used PL/SQL HTTP and XML support to communicate with other parts of the system from within the Oracle stored procedures. I developed SQL queries which were used for report running against the Impala database on top of the HBase data. I also developed Postgres stored procedures for data load.