Data Scientist, Lingokids

• Strong statistical foundation: deep expertise in probability, statistical inference, regression modeling, and causal inference methods (A/B testing, difference-in-differences, propensity score matching, uplift modeling). Knows when each tool is appropriate and where each breaks down
• Python for data science: fluent in Python (Pandas, Numpy, Scipy, Statsmodels, Scikit-learn) for analysis, modeling, and prototyping. Writes clean, reproducible, version-controlled code
• SQL and data pipelines: strong SQL skills for large-scale data warehouses (Databricks or equivalent). Comfortable with DBT or similar transformation layers to build robust analytical foundations
• Experimentation expertise: has designed and run A/B tests end to end in a product context, including power analysis, metric selection, and handling tricky scenarios (novelty effects, network interference, multi-armed bandits)
• Recommendation and personalization thinking: experience evaluating or improving how a system decides what content, products, or actions to surface to users. Understands the core trade-offs in personalization - relevance vs. diversity, exploration vs. exploitation, short-term engagement vs. long-term value - and has worked on measuring whether recommendations are actually helping users, not just optimizing a metric
• Complex problem-solving: comfortable sitting with ambiguous, open-ended problems and structuring a rigorous path to an answer. Does not default to simple solutions when the problem demands more
• Product and business thinking: connects statistical work to business outcomes. Asks the right question before answering the one given. Has influenced product or personalization strategy with data-backed recommendations
• Collaboration: works effectively at the intersection of analytics, product, and engineering. Translates statistical complexity into clear implications without oversimplifying
• English is a must: We’re a multicultural team providing a service in English, so while certifications aren’t necessary, fluency is essential. As a fully remote company, clear and effective spoken and written communication, especially in asynchronous and long-form formats, is key to collaborating successfully

Desirable

• As a Data Scientist on the Product Engagement team, your mission is to push our recommendation and personalization capabilities further by solving the complex statistical and methodological problems that sit beyond standard analytics
• Where the analyst maps the user journey and identifies opportunities, you are the person who figures out how to act on them at scale - designing the statistical frameworks, causal models, and experimentation strategies that turn behavioral insights into smarter, more effective personalization
• You thrive on hard problems, bring rigorous thinking, and are equally comfortable working with product managers and with ML engineers
• Advance our recommendation and personalization logic by designing and validating new statistical approaches - from improved segmentation models and engagement scoring to causal uplift estimation and multi-touch attribution
• Own the experimentation framework for the team: design rigorous A/B and multivariate tests, develop methodologies to handle complex measurement challenges (novelty effects, interference, long-term effects), and ensure results are interpreted correctly
• Tackle complex analytical problems that require going beyond descriptive statistics - building predictive models, running causal inference analyses, and decomposing metric movements into their drivers
• Partner closely with the Senior Analyst to turn user journey insights into statistically grounded targeting and personalization strategies
• Collaborate with the ML Engineering Data Scientist to ensure that statistical models and experimental findings can be produced and integrated into the recommendation engine
• Define and refine the metrics and success criteria used to evaluate personalization quality, ensuring we measure what actually matters for users and for the business
• Own the offline evaluation framework: build and maintain counterfactual evaluation methods, replay-based policy estimation, and offline screening pipelines that let the team assess 10 ideas offline for every 1 tested live - dramatically accelerating iteration speed
• Map and measure the interplay between editorial rules and personalization logic - systematically investigating when and how rules like hardcoded IP placements, content diversity constraints, or curated onboarding slots interact with ML-optimized recommendations, and designing experiments to quantify the trade-off
• Communicate complex statistical findings clearly and confidently to non-technical stakeholders, and advocate for methodological rigor across the team