Amazon
Applied Scientist
$172K–$222K
San Francisco, CA · Full-time · Posted Oct 8, 2026
- Experience
- 5+ years of hands-on industry applied ML experience with a Bachelor's, or 3+ years of applied ML experience with a Master's, or 1+ year of applied ML experience with a PhD; 1+ years in GTM / Content / Creator organizations is preferred.
- Education
- Degree in a quantitative field (Computer Science, Mathematics, Statistics, Physics, Engineering, or similar).
About the job
Applies machine learning to creator and content-context problems, developing models and products for Twitch’s Ads, Brand Safety, Creator Sponsorship, and Content Understanding areas.
What you'll do
- Build content-understanding ML products and lead high-impact projects from customer problem definition through execution.
- Collaborate with engineering and product teams to prototype and deploy models in data pipelines and ML services.
- Experiment with techniques in LLMs, computer vision, inference, and model evaluation.
What they ask for
- Experience partnering with technical and non-technical business owners to deliver results.
- Quantitative degree plus applied ML experience: 5+ years with a Bachelor’s, 3+ years with a Master’s, or 1+ year with a PhD.
- Experience with LLM model evaluations and owning projects involving numerous scientists and engineers.
- Expertise in Python, SQL, modern open-source ML libraries, and software development in AWS or similar cloud services.
Nice to have
- PhD specializing in Machine Learning; experience with computer vision, LLM inference, or content understanding.
- 1+ years in GTM / Content / Creator organizations, or familiarity with Twitch and its community.
Pay
171,600.00 - 222,200.00 USD annually (base salary)
Tools
Python, SQL, Pytorch, Tensorflow, AWS, LLM, computer vision
AI focus
Develops and deploys applied ML for content understanding, creator and viewer communities, Ads, Brand Safety, and Creator Sponsorship; work includes LLM evaluation and experimentation with LLMs, computer vision, inference, and model evaluation.
Model-written summary · the original posting is authoritative