• Led data science, machine learning, and analytical initiatives at a fast-growing B2B marketplace, collaborating with technical and business stakeholders to define strategic roadmap• Proposed and generated product embeddings, from images and text via pre-trained DL models, as a solution to known pain points – Vector (semantic) search POC, eliminated search results with zero/low number of matches on a large test set – Similar products recommendations, retailer taste profile (ML-powered recommendations based on user’s activity)• Identified possible improvements of our ML-powered ranking system (feature engineering, user behavior analyses) – Cumulative 25% order conversion uplift in A/B tests• Led critical analyses to evaluate the causal impact of partnership deals on incremental and shared GMV, as well as determining category incremental value to inform sales team focus

• Designed personalized ML-powered rankings, cooperated with Marketplace&Product teams on feature engineering – 11.5% order conversion uplift in A/B tests• Developed SQL and Python data pipelines orchestrated by Airflow. Co-built our data warehouse from scratch using dbt and BigQuery• Analysed user search behavior and re-designed our Elasticsearch-based search, co-created search improvement roadmap• Increased visibility, with Marketplace team, of best brands and products leading to average 25% increase in traffic & 10% in conversion

• Demand and elasticity modeling for pricing platform to determine optimal prices for FMCG (both offline and online) – Typically 3+% of additional profit/revenue with full explainability of every price• Identified and resolved platform limitations to onboard new clients - sparse sales modeling, missing promo detection model. Improved our sales prediction by detecting covid-related sales anomalies. Implemented product cannibalization model• Implemented uplift modeling – impact of promotions on sales both as uplift for the product itself and its cannibalization impact• Advocated for and spearheaded best SWE practices into our (R) codebase, e.g. unit tests, single responsibility principle, functionalitydecoupling and OOP, leading to better maintainability and faster iteration cycles• Co-designed with Product ”bulk pricing” module and created its data insights (”alerting”) functionality to increase users’ control

• Developed and optimized deep learning models for an algorithmic cryptocurrency trading system. Contributed to improving trading strategies• Implemented analysis and reporting automation tools, including a model comparison & evaluation framework to help pick best NN• Created data pipelines to extract and process data from multiple cryptocurrency exchanges’ APIs, automated data quality checks• Abnormal ECG beat detection – EDA, benchmark models, data preprocessing and PyTorch training pipeline

Part of team developing SQL reporting and calculation tool for IFRS9, predictive modeling of macroeconomic variables for credit risk model in R, validation of IFRS9 models

Data analysis & visualizations, senior staff support, MS Excel & Power BI