Overview As an Applied Scientist at Viant you will develop and productionize reinforcement learning and related decision-making models for real-time ad decisions. You'll work across contextual bandits, ranking, and bidding to optimize campaigns and auction efficiency at scale. Collaborating with engineers and product teams, you translate research ideas into reliable, low-latency models in a high-throughput advertising platform. Your work drives measurable business outcomes through rigorous experimentation and deployment in production.
Compensation / Benefits- fully paid health insurance
- paid parental leave
- unlimited PTO
Responsibilities- Develop, train, and evaluate RL, contextual bandit, ranking, and prediction models for ad optimization and targeting
- Study auction dynamics, budget pacing, and reward design to improve real-time decisions
- Translate research ideas into production-ready models for high throughput and low latency
- Design offline and online experiments including counterfactual/off-policy evaluation to measure impact
- Collaborate with engineers to deploy, monitor, and retrain models in production
- Apply quantitative methods to CTR, CPA, ROAS, attribution, and measurement problems
- Define objectives, labels, loss/reward signals, evaluation metrics, and experimental plans with cross-functional partners
- Contribute to a rigorous research-oriented culture through reviews and knowledge sharing
Key requirements- 1-3 years of ML model development with measurable outcomes
- Strong ML foundation with Python and PyTorch or TensorFlow
- Experience with reinforcement learning, contextual bandits, sequential decision-making, or related methods
- Ability to formulate machine learning problems with clear objectives, labels, features, and evaluation
- Experience analyzing large-scale data and communicating findings to scientists and engineers
- Interest in production-ready models that improve real-world decisions
- Experience with RL in advertising, marketplaces, or sequential decision-making environments
- Exposure to contextual bandits, counterfactual evaluation, causal inference, auction theory, or online experiments
- clear scientific communication
- cross-functional collaboration
- ability to explain technical concepts to non-technical partners
- Python
- PyTorch
- TensorFlow
