Description

Amazon is seeking a senior natural language processing engineer to own the Japanese language experience for Alexa. This hybrid product and language technology role combines product strategy ownership with hands-on language engineering expertise to build and localize GenAI technology for the Japan market. As the single-threaded owner of the Japanese language capability bar, you will define the vision, roadmap, and success metrics while providing direct technical and linguistic support to global teams across ASR, TTS, NLU, and conversational AI. This high-visibility role impacts millions of Japanese customers and establishes Amazon's AI voice leadership in Japan.

You will shape the future of Japanese AI voice technology at the intersection of product strategy and cutting-edge language technology. The ideal candidate is equally comfortable writing a product vision narrative for senior leadership as designing test sets to measure Japanese pitch accent accuracy, diagnosing error patterns, or automating evaluation pipelines. You can conduct deep dive analysis on language performance and bring compelling data to motivate change. You will work across Applied Science, Engineering, QA, UX, and business stakeholders in a fast-paced, ambiguous environment, structuring problems into actionable frameworks that drive measurable outcomes.

Key job responsibilities
Product Strategy & Ownership
• Define and own the product vision, strategy, roadmap, and success metrics for Japanese language capabilities in Alexa, including competitive benchmarking
• Drive product discussions and executive communication in both Japanese and English; bridge Japanese market needs and global technical capabilities, ensuring cultural and linguistic complexities are addressed in product development
• Influence cross-functional roadmaps and engineering priorities through data-driven contributions; make smart trade-offs across initiatives, balancing short-term delivery against long-term strategic goals
• Own end-to-end launch execution and post-launch quality monitoring, defining showstoppers and ensuring issues are triaged and resolved in priority order

Language Technology & Data Expertise
• Design evaluation test sets, define quality metrics, and establish regression testing and benchmarking methodologies for Japanese language performance across key user journeys, in partnership with QA and science teams
• Produce, process, and analyze language data to diagnose quality issues and inform product and modeling decisions; automate evaluation and data workflows using Python and/or internal NLP tooling
• Partner with Applied Scientists on training data requirements and the customer impact of architectural and data decisions for Japanese; provide Japanese language engineering support to global teams including data collection design, annotation guideline authoring, quality auditing, and model evaluation
• Identify and proactively communicate pitfalls unique to Japanese language and speech technology (e.g., homograph and homophone disambiguation, pitch accent assignment, argument and topic omission, appropriate keigo use in response generation) and develop mitigation strategies


A day in the life
Your morning kicks off designing an evaluation taxonomy for Japanese entertainment use cases with Applied Scientists. You map real utterance patterns against failure modes the model struggles with and ensure statistical coverage that catches real problems, not just easy ones. Next, you dig into a model benchmarking exercise with the QA team, comparing candidate models across performance metrics including Japanese-specific signals like pitch accent. By late afternoon, you're writing a technical explainer for a US engineering team, walking them through how Japanese orthographic complexity and compounding homophone ambiguity create failure modes they'll never see in English.