Software Engineer - Science Platform; BE

The Role

You'll build and maintain the platform that powers how the world's leading brands measure the true causal impact of their marketing spend. Our Science Platform runs geo-based experiments across 100+ customers, processes daily analysis pipelines, and delivers statistical results that directly drive budget decisions worth millions of dollars.

This is a backend and platform engineering role. You'll work primarily in Python across a set of tightly integrated repositories: a statistical estimation library, a science orchestration library, and a Metaflow-based job execution system running on Kubernetes. You'll collaborate closely with applied scientists to translate research into production code, and with product engineers to ensure results flow cleanly into the customer-facing application.

You won't be starting from a blank canvas — you'll be joining a production system that serves real customers and shipping improvements that compound. The engineers who thrive here are the ones who can navigate a complex, multi-repo codebase, understand the science well enough to be a productive partner, and ship reliable systems without needing to rewrite everything first.

We're also a team that leans into AI-assisted development as a genuine force multiplier. Our engineers use tools like Claude Code and Cursor to move faster, accelerate exploratory work, and ship features that would have taken weeks in days. We're looking for someone who's excited about this way of working.

• Build and evolve the data pipelines that fetch, aggregate, and transform KPI data from Big Query across multiple geographies and granularities

• Extend and maintain the statistical estimation library — implement new estimators, improve standard error methods, and optimize performance for large panel datasets

• Improve the Metaflow-based analysis orchestration system that schedules and executes thousands of daily experiment analyses on Kubernetes

• Design for reliability: build monitoring, alerting, and self-healing patterns for pipelines that run autonomously every day

• Collaborate closely with applied scientists to translate research prototypes into production-grade code with proper testing, error handling, and observability

• Work with product engineers to ensure analysis results are published correctly and flow cleanly into the customer-facing API and frontend

• Use AI development tools as part of your daily workflow to accelerate delivery and explore solutions

• Participate in on-call rotation and own the operational health of the science platform systems

• 3+ years of experience building and shipping production software systems

• Must have strong Python proficiency — you write clean, well-tested Python and are comfortable with the ecosystem (pandas, numpy, pytest, poetry)

• Experience with data-intensive applications: you've worked with large datasets, data pipelines, or ETL systems and understand the tradeoffs

• Experience with SQL and analytical databases (Big Query, Snowflake, or similar) — you can write performant queries and understand how warehouse-scale data processing works

• Comfort with cloud-native environments (GCP preferred): you understand how to deploy, monitor, and operate services in production

• Experience with workflow orchestration frameworks (Metaflow, Airflow, Dagster, Prefect, or similar) is a strong plus

• Track record of working effectively with AI development tools (Claude, Cursor, Copilot, or similar) — you've integrated them into your workflow and can articulate how they change the way you build software

• Ability to collaborate productively with scientists and researchers — you don't need a PhD, but you should be comfortable reading statistical code, understanding experimental design concepts, and asking good questions

• Excellent communication skills — you can explain technical tradeoffs clearly and work effectively across disciplines

• Earlier stage startup experience

• Familiarity with statistical or scientific computing (scipy, scikit-learn, Bayesian methods) — enough to be a productive partner to scientists

• Experience with Kubernetes and containerized workloads

• Experience with event-driven architectures…