Sr. Applied Scientist

What You'll Do

Viant's Machine Learning team is building autonomous advertising systems that make real-time decisions across targeting, ad optimization, bidding, measurement, and personalization. These systems process hundreds of millions of events daily and operate in the high-throughput, low-latency environment of programmatic advertising.

As an Applied Scientist, you will apply reinforcement learning and related decision-making methods to improve how Viant selects, ranks, and bids on advertising opportunities. You will work across contextual bandits, exploration and exploitation, counterfactual learning, and model-based experimentation to turn research into production systems that improve campaign performance, auction efficiency, and measurable business outcomes.

The Day-to-day

• Develop, train, and evaluate reinforcement learning, contextual bandit, ranking, and prediction models for ad optimization, bid optimization, targeting, and personalization.
• Study auction dynamics, delayed feedback, exploration and exploitation, budget constraints, pacing, and reward design to improve real-time advertising decisions.
• Translate research ideas into production-ready models that operate reliably at high throughput and low latency across Viant's advertising platform.
• Design and analyze offline and online experiments, including counterfactual and off-policy evaluation where appropriate, to measure model quality and incremental business impact.
• Partner with engineers to deploy, monitor, retrain, and improve models in production, addressing issues such as data leakage, class imbalance, drift, calibration, and changing market conditions.
• Apply quantitative reasoning and statistical modeling to problems involving click-through rate, conversion, return on ad spend, targeting, attribution, identity, and measurement.
• Collaborate with scientists, engineers, and product partners to define objectives, labels, loss functions, reward signals, evaluation metrics, and practical delivery plans.
• Contribute to a rigorous, research-oriented team culture through technical communication, code and model reviews, experimentation, and knowledge sharing.

Must Have

• 3-5 years of experience developing and applying machine learning models, ideally in production or research environments with measurable outcomes.
• Strong foundation in machine learning, deep learning, probability, statistics, and optimization, with practical experience using Python and frameworks such as PyTorch or TensorFlow.
• Coursework, research, internship, or project experience with reinforcement learning, contextual bandits, sequential decision-making, recommendation systems, online experimentation, or related methods.
• Ability to formulate a machine learning problem precisely, including objectives, labels, features, loss or reward functions, evaluation metrics, and experimental design.
• Experience analyzing large-scale data and communicating technical findings clearly to scientists, engineers, and cross-functional partners.
• Interest in building models that move beyond offline accuracy and improve real-world decisions in production systems.
• Experience with reinforcement learning in advertising, marketplaces, recommendation systems, robotics, games, or other sequential decision-making environments.
• Exposure to contextual bandits, off-policy or counterfactual evaluation, causal inference, auction theory, or online experimentation.

Great to Have

• Experience with digital advertising, real-time bidding, audience modeling, ad ranking, personalization, or large-scale recommendation systems is a plus.
• Experience with distributed computing, cloud platforms, LLMs, generative AI, or multimodal AI is also welcome, but the core focus of this role is production-oriented reinforcement learning and decisioning.