The company is an early-stage ad technology business with a small engineering team.
The team decides in real time which ads reach users. You will help build and improve those systems across data pipelines, machine learning and production serving, and your work will directly affect the recommendation system and the business results it drives.
Responsibilities
Build low-latency retrieval, ranking and reranking for ads, weighing advertiser performance against user experience.
Create generative agents that make playable and video ad variations, and send their performance data back into the recommendation system.
Build serving infrastructure for sub-second response times while keeping costs efficient.
Model users and their context from conversation data, engagement history, device, location and session signals.
Build pipelines and feature infrastructure to train models on reward signals that arrive at different times, from clicks to conversions.
Requirements
You have 0 to 3 years of experience and have shipped at least one ML or AI system to production.
You have worked end to end across backend systems, data pipelines, model training and production serving.
Ideally, you have built a recommendation or ranking system.
You have strong systems fundamentals and can reason from first principles about large-scale infrastructure, classical machine learning and AI agents.
Experience with PyTorch is relevant.
AdTech or recommendation systems experience is a plus.
You have an active GitHub or shipped side projects showing backend, ML or infrastructure work.
Interest in games and playable ad formats is valued. Startup experience, mentoring or building AI-native products is also relevant.
You must be a US citizen or green card holder and able to work on-site in San Francisco 5 days per week.
Conditions
Base salary is $200,000 to $250,000 per year, plus equity.
Visa sponsorship is not available.
The role is on-site in San Francisco, California, United States, 5 days per week.
Where you’d work
From the office, 4 or more days a week in the office