Andrew Russell G.

I’m a data scientist and mixed-methods researcher with a PhD in anthropology. I’ll help you understand anything about people: customers, market segments, social networks, trends, communities. Things I can do for you: • Design, carry out and analyze surveys. • Create people-centric KPIs and metrics. • Discover trends using AI-based textual analysis and ML models. • Fit predictive models using heterogeneous data. • Interview customers to deeply understand how and why they use your product. • Create data pipelines, AI prototypes and dashboards. • Communicate insights with slide decks, video presentations and reports. Selected results: • Modeled how Wikipedia articles in 40 languages develop over time (hidden Markov model, clustering, stratified sample of 20,000 articles), revealing patterns such as contested, abandoned and bot-built articles. • Traced the root cause of missed banner impressions in a live Wikipedia fundraising campaign. In Ireland, mobile impressions had fallen to half of pageviews. Segmenting by region, IP address and browser revealed the culprit: Facebook's Free Basics proxy, which never showed banners. • Mapped what readers in each world region read on Wikipedia, combining pageview data with demographic indicators and machine-learning predictions. Key insight: readers in almost every region lean toward topics related to their own region. • Interviewed 27 Wikipedia volunteer editors in 13 countries, caught a sampling effect by checking the findings against platform data, and delivered 12 product recommendations on a new feature. • Designed usage metrics for Wikidata, Wikimedia's knowledge graph, by analyzing the source code of millions of wiki pages with Spark SQL. Delivered 8 insights and 6 product recommendations. Skills: • Data wrangling: SQL, Spark, Trino/Presto, Hive, Druid, MySQL/MariaDB, Pandas. • Data visualization and dashboards: Matplotlib, Seaborn, Altair, Superset, Panel, Power BI. • Statistics and ML: frequentist and Bayesian modeling, HMMs, GMMs, state space models, relational event models, Prophet. • AI: LangChain, LlamaIndex, retrieval-augmented generation (RAG), fine-tuning, text classification. • Programming: Python, JavaScript, PHP, Jupyter, Git, Bash. • Qualitative: thematic analysis, grounded theory, interviews, focus groups, fieldwork. • Methods and methodology: A/B tests, experiment design, research design, conceptual frameworks. Background: • I have over 20 years of experience across quantitative and qualitative research, data science and software engineering. • I spent a decade at the Wikimedia Foundation, the non-profit responsible for Wikipedia, first as a software engineer with data science activities (time-sensitive analyses of fundraising banners, data collection and pipelines for A/B tests, seasonality models), and then as a research contractor. • Before that, I worked for a decade at the Instituto Mora, one of Mexico’s premier social science research institutes. • PhD in social anthropology, refereed research at Wiki Workshop 2025 and 2026, co-author of an award-winning textbook on qualitative research methodology. Remote since 2013, native-level English, Spanish and French. Tell me what you're trying to figure out, and I'll suggest an approach.