Samba is a media intelligence company. We know what the world is watching, reading, and thinking about — in real time, at scale, across every screen. Our data exists with the consent of over a billion people, organized into the most complete picture of consumer attention ever built. The biggest brands in the world use that picture to make smarter decisions. We think it’s the most interesting data asset on the planet, because it’s the most culturally relevant.
ABOUT THE ROLE
We are looking for a hands-on Data Scientist to own and deliver complex measurement science and modeling work at the core of our measurement and audience sciences products.
The role requires a deep, first-principles understanding of data science and machine learning — not just the ability to apply libraries, but the ability to reason clearly about model behavior, articulate trade-offs between approaches, and make defensible methodological decisions under ambiguity. This is emphatically a coding role — you will spend the majority of your time writing production-quality Python, building and evaluating models on large-scale viewership and web data, and delivering end-to-end ML solutions.
You will work closely with Data Engineering, Product, and go-to-market teams.
WHAT YOU'LL DO
-
Write and own production-quality Python code end-to-end — well-structured, tested, documented, and built to last; PySpark proficiency is essential for working with Samba's billion-row viewership datasets
-
Design, build, and deploy measurement models and statistical frameworks that power Samba’s campaign measurement, reach/frequency estimation, and cross-platform attribution products
-
Apply the right statistical and ML technique to the right problem — drawing from hierarchical models, Bayesian inference, gradient boosting, regularized regression, causal ML, and probabilistic record linkage — and clearly articulate the reasoning behind your choices
-
Build and evaluate multi-touch and multi-channel attribution models; apply Causal ML methods — counterfactual modeling, meta-learners (S-learner, T-learner, X-learner), and heterogeneous treatment effect estimation — to advertising and viewership measurement problems
-
Partner with Data Engineering to define data requirements, validate pipelines, and ensure model inputs are reliable, scalable, and production-ready
-
Lead technical design reviews and contribute meaningfully to architecture decisions across the Data Science team
-
Mentor junior Data Scientists through code review, pairing, and structured technical feedback — raising the team's technical floor
-
Communicate measurement methodologies and findings clearly to technical and non-technical audiences, including senior leadership and external clients
WHO ARE YOU
-
5-7 years of professional data science experience — hands-on, delivery-focused, and measurable in shipped models and production systems
-
Expert-level Python — clean, modular, testable, production-ready code is your standard, not your aspiration
-
Advanced PySpark and Databricks — comfortable building and optimizing data pipelines and ML workflows on billion-row datasets
-
Deep, first-principles command of statistics and ML — you can explain from the ground up how these models work and you apply this understanding to make better modeling decisions
-
Solid grasp of experimental design — A/B testing, randomization, power analysis, and the conditions under which observational causal inference is appropriate
-
Fluent in the full ML lifecycle: feature engineering, model evaluation, deployment pipelines, drift monitoring, and iterative improvement in production
-
Hands-on experience with uplift modeling, synthetic control, difference-in-differences, or propensity-based approaches applied to advertising or media outcomes
-
Strong ownership mindset — you drive projects independently and are comfortable owning your models from data exploration through production delivery, with minimal hand-holding.
-
Clear communicator — able to translate statistical reasoning and model behavior into language that drives decisions with product, engineering, and leadership
-
Experience with multi-touch attribution (MTA) or multi-channel attribution modeling — understanding of the limitations of rule-based approaches and the methodological trade-offs of data-driven alternatives
-
Hands-on experience with Causal ML methods — counterfactual modeling, meta-learners, and heterogeneous treatment effect estimation — applied to advertising or media measurement outcomes
-
Direct exposure to TV or digital viewership data — ACR signals, STB data, viewership panels, or cross-platform measurement (linear + CTV/OTT)
-
Familiarity with the measurement
-
t vendor landscape (Nielsen, Comscore, VideoAmp, iSpot) and industry standards (MRC, GRP/TRP frameworks)
-
Advanced degree (MS or PhD) in Statistics, Mathematics, Computer Science, or a related quantitative field — or equivalent depth demonstrated through work
