AI Engineer

Wayy Research is advancing the precision of measurement in digital marketing analytics. As an AI Engineer, you will own the algorithms for budget allocation, attribution, and rigorous experimental measurement, determining whether cross-channel effects are real or simply noise, all in a remote-friendly setting.

Responsibilities

• Develop and maintain a multi-armed bandit engine using Thompson Sampling on Beta-Bernoulli reward models, with contextual extensions as needed. Manage exploration-exploitation tuning, cold-start handling, and non-stationary reward drift.
• Build an attribution system around Shapley value-based multi-touch credit assignment, benchmarking against first-touch, last-touch, and position-based baselines.
• Design and operate experiment infrastructure for A/B and multivariate tests, including power analyses, sequential testing with alpha spending, and significance testing that remains valid despite peeking.
• Execute cross-channel causal measurement using geo holdouts, incrementality tests, and difference-in-differences designs to isolate true interaction effects from mere correlations.

Requirements

• Strong Python skills with experience in NumPy, SciPy, pandas, and scikit-learn.
• Hands-on experience with multi-armed bandits—Thompson Sampling, UCB, and epsilon-greedy—and a solid understanding of regret bounds, not just APIs.
• Expertise in hypothesis testing and experimental design, including power analysis, multiple comparisons correction, sequential testing, and addressing the peeking problem.
• Proven track record shipping production-grade A/B tests, not solely performing analysis.
• Causal inference experience—DiD, synthetic control, instrumental variables, or incrementality testing.
• Advanced SQL proficiency with PostgreSQL.

Technologies

• Python
• NumPy
• SciPy
• pandas
• scikit-learn
• PostgreSQL
• PyMC
• Stan
• NumPyro

Benefits

• Pay: USD 60.00 - 70.00 per hour
• Remote

You’ll fit if

You are uncomfortable with lift numbers reported without a control group and can articulate the difference between a channel that converts and a channel that causes conversions to a CMO without condescension.

Strongly preferred

• Marketing mix modeling or media measurement background
• Bayesian methods experience with PyMC, Stan, or NumPyro
• Shapley values or cooperative game theory applied to attribution
• Production ML deployment experience, not just notebooks