Applied ML Scientist Intern - RL & Optimization

Amazon; Palo Alto (CA); On-site; USD 115,000 - 156,000; Full time

Job summary

Amazon Science seeks visionary graduate student scientists for a 2027 Applied Science Internship in Reinforcement Learning & Optimization. Work on cutting-edge ML methods, deploy novel RL algorithms to real-world challenges, and collaborate with top researchers and engineers across multiple sites including Palo Alto.

This full-time internship offers access to state-of-the-art infrastructure and mentorship. Ideal candidates are PhD students with strong programming in Java/Python, PyTorch or JAX,

Qualifications

• Enrolled in a PhD program.
• Willing to relocate to internship location.
• Experience programming in Java, Python with PyTorch or JAX.
• Experience in Optimization, RL, LLMs, Time Series, Graph Modeling, DL, Predictive Modeling.
• Demonstrated depth via publications or thesis work.
• Available for full-time (40 h/wk) internship.

Responsibilities

• Develop and implement scalable RL-optimization algorithms for production-scale data.
• Collaborate with product managers, scientists, and software engineers to advance state-of-the-art.
• Communicate research findings clearly and contribute to publishing and reviews.
• Leverage AI-powered tools to accelerate research, and critically validate outputs.

Skills

Java

Python

PyTorch

JAX

Optimization

Reinforcement Learning

Time Series

Graph Modeling

Deep Learning

Statistical Modeling

Education

PhD student

Tools

PyTorch

JAX

Job description

Amazon Science seeks visionary graduate student scientists for a 2027 Applied Science Internship in Reinforcement Learning & Optimization. Work on cutting-edge ML methods, deploy novel RL algorithms to real-world challenges, and collaborate with top researchers and engineers across multiple sites including Palo Alto.

This full-time internship offers access to state-of-the-art infrastructure and mentorship. Ideal candidates are PhD students with strong programming in Java/Python, PyTorch or JAX,