Mariya Mansurova

Personal blog on Towards Data Science, mostly focused on GenAI topics

- Building an analytics team from the ground up in a new domain - Developed an ROI framework for the team’s focus areas (including challenging topics like the impact of standardisation) and defined key KPIs - Led analytics for the addresses domain: defined success metrics, built an ML model to predict address quality, and developed an LLM-based prototype for address validation

- Led a team of up to five analysts - Championed the adoption of GenAI within the analytics organisation, leveraging LLMs both in product development and for generating insights - Developed prioritisation frameworks that helped shape team strategy, including the Perfect Transfer Score framework, a North Star metric for the account product focus, and a customer impact model for Direct Access integrations. - Built prototypes for LLM-powered automation in the business onboarding flow, resulting in a 1.5x increase in automation rates and significant cost savings - Owned analytical prioritisation for a major repapering campaign, ensuring compliance with regulatory deadlines despite limited operations capacity and with minimal customer impact.

Product Analytics for Yandex.Metrica (the #3 web analytics platform globally according to W3Tech) and AppMetrica (one of Russia's leading mobile analytics platforms) Key responsibilities: - Managed a team of up to 4 analysts - Led the analytical tools development stream (creating intelligent solutions for customers) within Yandex.Metrica - Handled product management for Yandex.Radar service (providing high-level internet statistics) - Analysed big data using various tools including Python, pandas, SQL (ClickHouse, MySQL), and MapReduce to support data-driven product decisions Major projects: - Designed the recommendation system architecture for Yandex.Metrica (including new feature promotions, anomaly detection, and advisory insights for better data utilisation) - Leveraged Yandex hackathons to gather a team and implement product features within 2 weeks: 1) auto-conversion tracking for payment system returns 2) experimental reporting framework (enabling quick testing of new visualisations with minimal development overhead) - Established monitoring systems for product KPIs and data integrity to minimise service incidents

Key responsibilities: - Analysed massive datasets (>10 PiBs) using SQL (primarily ClickHouse), MapReduce, and Python - Created visualisations and dashboards with DataLens (Yandex's BI platform) and Python libraries such as Plotly - Conducted technical analytics including resource usage forecasting, code version validation, and data monitoring - Participated in various industry conferences and meetups advocating for ClickHouse and Yandex.Metrica in analytical and marketing communities Major projects: - Developed an ML-based ranking system for all websites in the Russian internet - Case with one of our customers: Built a customer conversion prediction model used for bid optimisation that delivered a +96% increase in revenue

Key responsibilities: - Identified and investigated KPI metric anomalies for Yandex.Search and Yandex.Browser by analysing external data from web analytics platforms (including Yandex.Metrica, LiveInternet, TNS) and other sources (Twitter, news feeds, etc.), as well as internal data (processing logs via MapReduce and querying our ClickHouse database) - Developed regular reporting systems to detect and isolate anomalies - Built and maintained real-time monitoring and alerting infrastructure for KPI metrics

Projects: Integration with ICOMS Billing System for Bright House Networks, USA and COX Communications, USA - Designed APIs for integration (Order Management, Customer Management, and Billing Operations) - Supported development and testing phases - Created design presentations - Led a team of analysts (up to 5 members)