Applied Scientist

About the Role

As one of Amazon's first Applied Scientists on our new Bengaluru team in Kindle Direct Publishing (KDP), you will help turn innovative ideas into essential models for authors globally. You will build agentic solutions leveraging large language models (LLMs) and multimodal AI to support millions of authors.

Key Responsibilities

• Build Multi-Agent Systems: Develop the orchestration, personalization, and evaluation logic behind KDP's author support assistant using specialized agents and tools for continuous improvement.
• Multimodal Modeling: Create models of book attributes to power similar-book search across languages, enhance catalog quality, and drive category/keyword recommendations.
• Model Optimization: Post-train open-source models on KDP expertise using techniques like supervised fine-tuning (SFT) and reinforcement learning, deploying where they excel in accuracy, cost efficiency, or speed over general-purpose models.
• Cross-Functional Collaboration: Work with US-based scientists and engineering partners to take work into production. Design rigorous offline/online experiments and write clear readouts with recommendations.

We serve authors ranging from first-time writers to full-time publishers, ensuring safety before speed and building shared foundations.