Aleksandr Isakov Email & Phone Number

Moscow, Ru

Moscow, Russia

Analyzed multiple hypotheses to forecast the total number of orders.A short-term forecast model was developed and implemented in production - the forecast for 2 weeks ahead is calculated in granulation by hours for each store with daily frequencyA mid-term forecast model has been developed and implemented in production - the forecast for six months ahead is calculated for each city. This takes into account the company's marketing activities and factors that may affect demand.There is also a process that facilitates the validation of the forecast and the process of its approval by interested departments. As a result, the forecast is validated and used further for their own purposes.Also based on the forecast, a model for forecasting the number of orders and items in more than 200 defined segments has been implemented.Additionally:1. The forecast of ice consumption at the stores for 1.5 years in advance is calculated. 2. Considered scenarios of forecast behavior depending on external events.3. Application of the forecast as a benchmark for economic forecasts

Moscow, Russia

Worked with multiple datasets to perform analysis, researched various hypotheses for fraud reasons. Developed models for partial repayment, and probability of default prediction. Created an XML parser for data processing from the credit history bureau. Participated in the completed development cycle - from setting the task together with the customer to implementing the model in the production. Projects:Customer profit prediction - in progressOperational risk losses prediction - Worked with time-distributed data. Data were processed with the exception of extreme losses using the POT (Extreme Value Analysis) method. A stable model is constructed under conditions of turbulently changing input data. The model has been successfully implemented in production to calculate various scenarios for the development of events, what reduces uncertainty for decision making.Fraud detection - Various hypotheses on the identification of groups of clients with a high concentration of fraud are analyzed. Multiple hypotheses were proposed also. An interpreted model was implemented in production for risk-score calculation on big data.Model development for partial repayment prediction in the early collection - Developed the model for partial payment prediction. Data has 10M observations. A complete analysis of the features based on their correlation and contribution to the forecast is carried out. The model has been successfully built in production.Online-update research - Investigated how online updated features can influence the model performance for next best action (NBA task) prediction. Data has 1.6M observations. It is proved that in the case of an online update, the model can improve efficiency in used metrics and save >400M for the bank.Extracurricular activities:Expert on the Scoltech & Sber Data Science hackathonInterviewed candidates for the data scientist position in our team

Moscow, Russia

Information Systems and Technologies. Energy Systems Laboratory.The analysis of various ways of controlling indoor microclimate systems is carried out: from rule-based to AI-based control, what leads to minimal energy consumption for peak load reduction at highload power systems.An extensive analysis of international practices in demand response has been carried out. Modern domestic developments and programs in this area are also considered. A solution has been proposed for consumers in the Russian climate using thermal energy storage to participate in demand response programs. Additionally: Monitoring of microclimatic parameters in the room using the Waspmote device. Sensor installation and configuration were done. Also, data was transferred to the ThingSpeak cloud in real-time.

Moscow, Russia

I worked on the broad literature review of technological and management methods used in Demand Response. A comparison of ways to control the microclimate system was made.

Moscow, Russia

Working on a project: "Air drying using adsorbents". In the work process, we built laboratory installation for air drying research. Numerous studies have also been conducted on the drying efficiency for various mixtures of adsorbents. The methodology for the installation test was created. Conduction of experiments and analysis of the results obtained.

Master'S Degree, Information Systems And Technologies. Energy Systems, 4,75

Courses: Introduction to DS, Machine Learning, Optimization Methods, Leadership for innovators Thesis: Mathematical modeling.

M.Sc. In Cryogenic Systems, Hvac, Refrigeration And Cryogenic Systems, 4,6

Courses: Probability theory, Linear Algebra, Calculus, Differential equations, Applied physics Thesis: Optimization of the refrigerant.