Valeriy Kondruk

- Dynamic floors model. - Designed and launched Rewarded Ads data foundation reducing frictions in tracking and scaling up Roblox's ads business. - Created a framework that automates game-level ad incrementality analysis in statistically robust manner. - Launched a model to identify 'likely spenders' allowing game developers to balance IAP and Ad revenue effectively by reducing cannibalization. - Designed ads revenue forecasting powering developer-facing feature.

- Developed bid request optimization model improving DSPs performance and generating $4M+ per quarter. - Identified 100+ user-level features, created automated feature importance framework. - Initiated the project of ingesting first-party data and released a model with up to 15% lift in publishers metrics (revenue, CPM, ARPU). - Developed Contextual Bandits model for auction floors (min price) optimization. - Contributed to dynamic throttling model development, reducing serving costs by 70% and improving advertisers ROI. - Developed Bayesian A/B testing framework for unevenly allocated and small experiments. - Implemented new features for Neural Network market price model, increasing revenue by 1.5%. - Migrated market price model's ETL from Amazon Athena to Iceberg in Data Lake. - Helped develop impression-to-click model, replacing primary impression-to-install model for Apple SKAN traffic.

- Managed development of a Neural Network model that predicted ad auction market price values, improving bid shading quality and increasing revenue by 10%. - Collaborated with a team of four Data Scientists. - Contributed to data preparation, feature engineering, code review, and performance analysis. - Managed project timeline, resources, and reported results to leadership.

- Developed an ML model that improved main model's prediction accuracy, boosting annual net revenue by 3% (millions of dollars). - Created a Python library for Synthetic Difference-in-Differences causal inference analysis, now used by multiple teams to quantify experiment outcomes when A/B testing is inapplicable. - Designed a methodology to evenly partition test units across groups, ensuring parity in key metrics like revenue, views, users, and CTR, without costly simulations, handling thousands of units. - Contributed to ML-powered recommendation system utilizing similarities among video ad attributes (mood, primary character, action level) to match candidates. - Participated in video taxonomy project using AI and computer vision to convert video ads into structured data. - Developed ETL pipelines using Python and Spark to collect and analyze multi-source, multi-format data (data lake, SQL databases, AWS S3 parquet files, JSON, vendor logs).

- Developed new pre-filtering rules which reduced recommendation system candidate pool from 10k to under 1k, cutting serving costs by 50%. - Discovered the bug impacting banners performance - the fix led to a 20% revenue lift. - Communicated analytical findings to the leadership team, translating complex data into clear and actionable insights. - Executed dozens of multivariate and conventional A/B tests. - Employed robust statistical techniques pre- and post-experiment to validate the reliability and integrity of test results. - Built ML models including XGBoost and neural nets to demonstrate predictive power of diverse ad attributes like game genre, character persona, ad duration, and more. - Conducted time-series analyses with Facebook Prophet and Google Causal Impact libraries in R and Python. - Created and maintained version-controlled LookML models. Built dashboards in Looker.

Ascape VR was the leading supplier of immersive travel content for brands like Marriott, JetBlue, Lonely Planet, Yellow Pages, Thomas Cook, and Amadeus. Ascape VR mobile app was installed by over a million users globally. - Managed the team of 8. Led product design and development, product analytics, and user acquisition. - Achieved over a million installs on iOS and Android. - Increased paying users rate from an industry average of 1.2% to 7.5%. - Received the 'Most Innovative Android App' award from Google Play Store in 2017. - My day-to-day job included user engagement analytics, features A/B testing, product features planning and forecasting.

- Managed fund's US business. - Managed venture deals from scouting to investment.

- Performed product analytics tasks covering user acquisition, retention, and growth rates. Conducted high-level financial forecasting. - Tracked the performance of the portfolio companies. Built dashboards in Tableau.

- Performed financial analysis and modeling on real estate investment opportunities across Europe. - My analytical reports were often published in Forbes Kazakhstan and other finance-related media.