WHAT YOU'LL DO
The Machine Learning team at Viant builds the real-time decisioning systems that power programmatic advertising — models that predict, in milliseconds, whether an ad opportunity is worth bidding on, and systems that decide how much to bid, live, across billions of auctions.
We're looking for an experienced Director of Applied Science to own this technical domain end to end: the models, the production systems they run on, and the team that builds them. This is a hands-on, player-coach role — you'll review designs, unblock scientists on hard formulation problems, and still be able to work through the math and the system tradeoffs yourself when it matters.
THE DAY-TO-DAY
- Own the end-to-end design, deployment, and performance of Viant's real-time prediction and bid-optimization systems — models that run inside strict latency budgets and directly drive auction outcomes
- Define the technical formulation for ambiguous ad-optimization problems: what's being optimized (click, conversion, ROAS, or incremental value), who the valid training population is, what the model should predict, and how a downstream system should interpret it
- Provide technical and strategic leadership to a team of applied scientists and ML engineers — reviewing designs, coaching through production tradeoffs, and setting technical direction
- Own the operational realities of production ML at scale: delayed feedback and attribution, label imbalance, latency and serving constraints, monitoring, drift, and retraining
- Evaluate and improve the systems that decide how bids are placed and adjusted in live auctions, incorporating budget, volume, and business constraints into the model's output
- Collaborate with Product and Engineering leadership to align technical roadmap with business priorities
- Build toward more autonomous, multi-step decisioning systems as the team's technical roadmap expands beyond single-prediction models
MUST HAVE
- 10+ years of experience building and deploying production machine learning systems, with at least 5+ years in a hands-on technical leadership role
- Direct experience with real-time, high-scale decisioning systems — ad optimization, bidding, ranking, recommendation, or a closely comparable domain with similar latency and volume constraints
- Demonstrated ability to formulate an ambiguous business problem into a precise ML specification: labels, training population, objective, loss function, and evaluation metrics — without relying on framework-level generalities
- Working fluency in the operational demands of production ML: delayed feedback, latency and serving tradeoffs, monitoring, drift, and retraining
