Data Scientist

На данный момент Data Scientist.

Мои компетенции и опыт

I am a Data Scientist with 3 years of experience in machine learning (ML), AI (LLMs, Agents, RAG), statistical research and causal inference (A/B testing and quasi-experimental designs), SQL, Python and BI systems. 
I specialize in optimizing IT solutions through the design and power optimization of A/B tests, as well as conducting rigorous causal inference studies to identify growth opportunities in existing products and systems. My experience in selecting product metrics and building interactive BI dashboards enables stable and reliable monitoring of IT products.


Marketing Analyst, Avito

Feb нужен доступ к резюме May 2026

Built a dataset of hundreds of TV campaigns by linking external exposure data with post-campaign brand-lift research.

Developed an LLM-embedding clustering pipeline to group multiple ad variations into distinct creative-level entities, saving months of manual labeling effort.

Designed and optimized a survey experiment to estimate the effect of creative features on ad quality using mixed-effects residualization and creative-level regression.

Data Scientist, Sberbank Technologies
Jan нужен доступ к резюме Feb 2026

Led data analytics for the GigaCode AI assistant (40k40k MAU).

Proposed and refined an A/B testing strategy for comparing LLM models, reducing test duration by more than 2x2x through changes in the randomization approach, the inclusion of additional covariates, and the use of pre-experimental data (CUPED).

Maintained and optimized SQL scripts for ETL pipelines.

Partnered with stakeholders to identify reporting needs and pain points, and designed interactive dashboards in Qlik Sense and Metabase with built-in data cleansing, user segmentation, and outlier removal logic to reliably track engagement, feature adoption, retention, and task success.

Led the development of an LLM evaluation service ( нужен доступ к резюме + Python; React frontend) that runs benchmarks and organizes results for storage. The service supports side-by-side comparison of AI models with result visualizations segmented by competency area, accelerating quality gate checks for new model production releases by 2x2x.

Applied quasi-experimental methods (Diff-in-Diff and ITS) to regularly evaluate the effects of new releases on product metrics and to estimate the impact of the AI assistant on developer productivity within the bank.

Developed a Python library for automated A/B experiment validation on historical data, featuring CUPED, the delta method, and clustered standard errors, improving statistical power and capturing a 10%10% effect on a core metric previously undetected by offline benchmarks.

Trained a predictive ML model as part of a causal analysis of user churn.

Developed and published a robust evaluation pipeline for AI code assistants in single-line completion tasks, designing a custom Levenshtein-based quality metric and evaluating statistical significance via hierarchical bootstrapping: нужен доступ к резюме