Marketing data scientist for user acquisition

We're building AI agents for performance marketing. Plurio is the AI agent for media buying teams running $300K–$15M/month on Google, Meta, and TikTok. It connects to your internal backend data and your ad accounts, automates analysis and optimization, saves 50% of time, and boosts ROAS by 10–30%. You own forecasting in Plurio: the conversion and LTV forecasts our clients' ad budgets are optimized against.

• Forecast conversions and LTV from a user's first events and update them as new data arrives. Your inputs are the client's own data, ad-platform data, and our attribution.
• Statistics, ML, or LLM-based approaches: whatever beats the baseline forecast on the client's data. No favorite models.
• Build a reusable solution that launches quickly for a new client, then tune it to that client's funnel and data.
• Backtesting on historical data and validation on a time-based holdout; error and bias by horizon and by segment; calibration. You develop the methodology for evaluating forecast quality and keep tracking it in production.

• Production forecasting models with measurable impact. You've launched LTV, conversion, or churn models that the business relied on.
• Strong SQL and Python. We barely write code by hand — AI does the heavy lifting. Your job is to make sure it's correct.
• You can explain it to the business. You tell a marketer in plain words how accurate a forecast is and what it depends on.
• You work with AI every day and build with it, not just chat with it.
• Professional-level English: you'll communicate directly with client teams.
• Будет плюсом: Experience in marketing: performance marketing, user acquisition, attribution; ClickHouse, dbt

• Competitive compensation, aligned with senior data-science experience; specifics discussed individually.
• AI-first culture, for real: the whole company works in Cursor with a shared workspace and premium team access to AI tools. You'll have the best agentic tooling in the industry and a team that expects you to use it.
• Maximum impact: your models go live in client accounts, not into slide decks. No corporate bureaucracy between you and the result.