Lead Data Scientist, AdTech

Launch Potato•Dallas, OK

14d•Remote

About The Position

Launch Potato is hiring a hands-on individual contributor to build and operate data science models for its auto insurance marketplace and media buying. The role will develop models that determine which insurance carrier a shopper sees on the click wall, aiming to route each user to the carrier most likely to convert. It will also own buying models for media spend that are evaluated on return on ad spend (ROAS), putting carrier conversion data—including quotes, policies, and acquisition cost—to use in production. The work involves delayed outcomes that may arrive weeks or months after a click, sparse sessions, repeat shoppers, market drift, and feedback loops in which bidding decisions affect the data available for future model training. The scientist will frame problems with business partners, find and work with the data, build and validate models, and follow them through deployment and monitoring. An ML engineer supports production work. Auto insurance will be the initial focus, with scope expected to broaden to other verticals over time. The company is remote-first, and most employees work East Coast hours. Success includes independently running insurance marketplace media buying by day 60. By the end of the first year, expected outcomes include a click wall that matches users to suitable carriers, carrier conversion data used in production models, automated marketplace bidding scaled beyond proof of concept, a clear approach to repeat users and sparse sessions, robust production models that handle drift and changing objectives, and trusted output used by the business without rework.

Requirements

At least 5 years of hands-on applied data science experience with measurable business impact.
Experience building models that drove paid media decisions, such as spend, bidding, or ad serving; attribution and reporting alone do not qualify.
Ability to explain the models on one’s resume in depth, including loss-function and architecture choices, as well as neural network fundamentals.
Experience handling how current bids shape future training data, and experience with late labels, sparse features, or drift.
Experience owning models end to end, from finding data through building, deployment, and monitoring.
Expert Python and SQL skills.

Nice To Haves

Experience spending real ad budget in live auctions.
Experience with large-scale ranking or recommendation systems.
Ability to explain tradeoffs to non-technical partners in plain language.
Experience designing for auction feedback loops, selection bias, and exploration.
Hands-on neural network experience, such as embeddings of text, creative, or user behavior.
AWS data science experience with ECS, S3, Athena, or SageMaker.
CI/CD deployment experience.

Responsibilities

Build models to select which auto insurance carrier each shopper sees on the click wall, optimizing for carrier conversion.
Own media buying models and bidding decisions that are evaluated on ROAS.
Use carrier conversion data, including quotes, policies, and acquisition cost, in production models.
Address delayed labels, sparse sessions and features, repeat shoppers, market drift, and the effects of bidding decisions on future training data.
Frame data science problems with business partners and independently find and prepare the data needed to address them.
Build, validate, deploy, and monitor models end to end, partnering with an ML engineer on production.
Run insurance marketplace media buying independently by day 60.
Scale automated marketplace bidding from proof of concept and develop production approaches for repeat users, sparse sessions, drift, and changing objectives.
Broaden the initial auto insurance work to other verticals over time.