Senior Data Scientist, Growth AllianceActive

Work with HelloFresh in Warsaw and its HelloTech organisation, HelloFresh’s global technology backbone with more than 1000 people, building the digital products that power our end-to-end food experience. From meal kits and ready-to-eat meals to specialty offerings like pet food…

What you'll do

• Send Decisioning: Designing and scaling AI agents, specifically using contextual bandits, to optimize our communication channels (email, SMS, and push notifications). Your models will automate the logic that determines the most relevant message and timing for each individual user.

• Intelligent Funnel: Supporting a collaborative, cross-tribe initiative aimed at personalizing the post-click experience. You will leverage contextual bandits to match user ad-context and pre-questionnaire preferences with the optimal landing page journey.

• AI-Assisted Content & Search: Building intelligent systems to index and search internal marketing assets (images and copy). Additionally, you will develop tools that assist marketing teams in the automated composition of new, high-performing communication assets.

• Build, own, and iterate on RL and ML models that run in production behind: high-traffic, performance-critical conversion and communication flows.

• Develop the decisioning logic that determines the best next action for each: user across the funnel and the messaging lifecycle.

• Drive the architectural evolution of our contextual-bandit systems: from online exploration (Bootstrapped Thompson Sampling) toward offline policy evaluation and deterministic, low-latency inference (Counterfactual Risk Minimization, Offset Trees, model distillation).

What they're looking for

• Partner with Engineering on system design so models integrate cleanly into: the funnel and messaging architecture — including where models decide, how decisions are served at latency, and how failures are handled and rolled back.
• Own experimentation design for decisioning models: define success metrics and guardrails, and interpret results in high-noise, high-traffic environments using rigorous methods (sequential testing / SPRT, propensity handling, allocation-bias-free setups).
• Keep models safe and effective in production: monitoring performance, input-data reliability, drift, and latency, and iterating quickly when issues surface.
• Collaborate with CRM Ops agent leads, Marketing Analytics, and platform: partners to turn model behaviuor into decisions marketers trust.