Lead Data Scientist

About Beamable

Beamable is the backend and LiveOps infrastructure platform for game developers, the system that lets studios build, operate, and scale live games without constructing the underlying architecture themselves. Backend services, identity, analytics, meta-game systems, virtual economies, and AI-native tooling: all integrated into a single platform that eliminates the point-solution stitching that has defined game development for too long. Beamable also powers Wingman, an agentic LiveOps product that gives game teams a unified operating system to understand, optimize, and act on live game performance in real time. This is the future of how games are operated.

Beamable is part of FIRY, a founder-led global technology portfolio operating at the intersection of gaming, AI-powered performance marketing, and developer infrastructure. Acquired in January 2026, and together with Skillz, our competition platform, and RZR, our AI-driven ad network, Beamable sits at the foundation of a compounding ecosystem that no standalone competitor can replicate. If you want to build the infrastructure that games run on at a company with the vision and resources to back it — Beamable is the right place to do it.

About the Role 

As Beamable’s Lead Data Scientist, you’ll own the evaluation methodology for our predictive live-ops models (churn, lifetime value and purchase propensity) across our studio client base. You’ll design how live-ops interventions are measured, define the guardrails for our self-tuning game loops, and partner with Beamable engineers to ship trustworthy models into production. You’ll deliver clear, defensible results to Beamable executives and studio clients, and influence real product and business decisions.

Responsibilities 

• Own the evaluation methodology for all predictive live-ops models (churn, lifetime value, purchase propensity), including out-of-time backtesting, leakage checks, drift detection and promotion rules. Set the standard for what counts as a trustworthy, shippable result.
• Design and run the measurement framework for live-ops interventions, choosing the right method for each game (A/B tests, uplift models, bandits, constrained Bayesian optimization). Own go/no-go decisions on whether interventions are working.
• Define the objectives, guardrails, step limits and early-signal metrics for Beamable’s self-tuning game loops, including when tuning should proceed automatically and when a human decision is required.
• Track developments in tabular ML, survival analysis, causal inference, bandits and sequence models. Run rigorous head-to-head comparisons before recommending adoption.
• Write production-quality Python training and scoring jobs, and partner with Beamable engineers on pipelines and product surfaces.
• Deliver clear, defensible results to executives and studio clients, including negative and uncertain results. Translate model behavior into language non-technical studio operators can act on.

Key Competencies 

• Evaluation rigor and honest science: deep understanding of how ML evaluations fail (leakage, overfitting, distribution shift, baseline selection), and a track record of reporting results honestly, including when a model doesn’t beat the simple approach.
• Experimentation and causal inference: strong command of test design, power analysis, sequential testing, holdouts, uplift modeling and off-policy evaluation on messy, real-world behavioral data.
• Probabilistic modeling on tabular data: expert-level experience with gradient boosting, survival models, calibration and leak-free backtests. Comfortable with small, shifting datasets.
• Research judgment: reads new papers critically, designs fair comparisons, and assesses whether results will hold up on real studio data, not just in benchmark conditions.
• Production ML engineering: writes clean, production-quality Python (pandas, scikit-learn, LightGBM or similar) and SQL, and partners with engineers to ship models into live product surfaces.
• Senior stakeholder communication: proven ability to explain uncertain or negative results to senior, non-technical leaders and studio clients, and change real decisions as a result.

Experience 

• A Master’s degree in Computer Science, Machine Learning, Statistics, Biostatistics, Economics, Operations Research or a related quantitative field; PhD strongly preferred.
• 7+ years applying data science or ML in industry, including systems whose results changed real product decisions.
• Deep experience with experimentation and causal inference: test design, power analysis, sequential testing, holdouts, and how experiments mislead.
• Strong probabilistic modeling on tabular behavioral data: gradient boosting, calibration and leak-free backtests.
• Production-quality Python and SQL, with a record of shipping models into live product environments.
• Proven ability to explain uncertain or negative results to senior non-technical leaders and influence real product decisions.
• Preferred: experience with survival models, bandits, uplift modeling or Bayesian optimization.
• Preferred: free-to-play live-ops experience (retention, monetization, in-game economies).

Starting Base Compensation: $78,000 

Location: Remote in Brazil

Travel Requirement: Less than 10%

Beamable embraces diversity and is proud to be an equal opportunity employer. As part of our commitment to diversifying our workforce, we do not discriminate on the basis of age, race, sex, gender, gender identity, color, religion, national origin, sexual orientation, marital status, citizenship, veteran status, or disability status, and we operate in compliance with the San Francisco Fair Chance Ordinance.

#GC-Remote  #GC-1  #MN-Remote   #MN-1