Data Scientist

GCS Recruitment Specialists; California (MO); On-site; USD 140,000 - 200,000; Full time

Job summary

Amazon Translation Services is seeking a Data Scientist to lead GenAI and LLM initiatives, building production-grade models for multilingual content translation and localization. You will fine-tune models, optimize OCR and image processing workflows, and deploy scalable ML systems with low latency.

The role demands 3+ years of ML model experience, deep LLM expertise beyond prompts, and strong Python skills.

Qualifications

• 3+ years building machine learning models for real business applications.
• Hands-on LLM experience that goes beyond prompt engineering: fine-tune models, adjusting weights and parameters, and working directly with model internals.
• A solid understanding of how foundation models work under the hood, and the judgment to pick the right model for the problem. Open source or closed models are both fine.
• Experience building and deploying production-grade models at scale.
• Strong programming in Python. Java or C++ backgrounds are also welcome.
• PhD, or Master's degree plus 4+ years of experience in Computer Science, Computer Engineering, Machine Learning or a related field.
• Comfort with ambiguity and autonomy on an early-stage team, where engineering judgment matters as much as model performance.

Responsibilities

• Evaluate and select models for incoming text, image and video content.
• Fine-tune models across the full image localization workflow: OCR extraction, grouping text within images, translation, and rendering translated text back onto the image. The work spans many languages and content types.
• Match model size to the problem. That means choosing the smallest model that does the job well rather than over- or under-engineering it.
• Find automation opportunities across the translation pipeline (training, sampling, inferencing) and work with the MLOps team to automate repeated manual work.
• Design and ship production-grade ML solutions that handle millions of pieces of content daily at low latency.
• Run data analysis to drive recommendations, and explain technical approaches clearly to both technical and non-technical stakeholders.

Skills

Python

LLM fine-tuning

Foundation models

Production ML

Model deployment

Unix/Linux

Data analysis

Education

PhD or MS + 4+ years in CS/ML

Tools

MLOps

Job description

Data Scientist, GenAI / LLM

About the Program

Amazon Translation Services (TS) is the team that makes Amazon's content available in every language it sells in. It translates billions of words a year across 130+ locales for developers, product and program managers, and linguists across the company. Its mission is a hands-off-the-wheel translation service: content goes in, accurate translations come out, with minimal human touch and at the lowest possible cost.

You would join a new science team within TS that runs with a start-up mindset. The long-term technical strategy is still taking shape, so there's real room to design GenAI solutions from scratch. The models you build and deploy now will process millions of pieces of content every day.

The Role

This is a hands-on applied science role focused on production work, not pure research. You'll take large language models and multimodal models, fine-tune them for problems that no off-the-shelf model solves, and deploy them at scale with low latency. You'll work alongside software engineers, data scientists, product managers and a dedicated MLOps team.

What you'll do:

• Evaluate and select models for incoming text, image and video content. This includes working out whether a piece of content holds a translation opportunity, such as text embedded in an image.

• Fine-tune models across the full image localization workflow: OCR extraction, grouping text within images, translation, and rendering translated text back onto the image. The work spans many languages and content types.

• Match model size to the problem. That means choosing the smallest model that does the job well rather than over- or under-engineering it.

• Find automation opportunities across the translation pipeline (training, sampling, inferencing) and work with the MLOps team to automate repeated manual work.

• Design and ship production-grade ML solutions that handle millions of pieces of content daily at low latency.

• Run data analysis to drive recommendations, and explain technical approaches clearly to both technical and non-technical stakeholders.

What We're Looking For

Required:

• 3+ years building machine learning models for real business applications.

• Hands-on LLM experience that goes beyond prompt engineering: fine-tune models, adjusting weights and parameters, and working directly with model internals.

• A solid understanding of how foundation models work under the hood, and the judgment to pick the right model for the problem. Open source or closed models are both fine.

• Experience building and deploying production-grade models at scale.

• Strong programming in Python. Java or C++ backgrounds are also welcome.

• PhD, or Master's degree plus 4+ years of experience in Computer Science, Computer Engineering, Machine Learning or a related field.

• Comfort with ambiguity and autonomy on an early-stage team, where engineering judgment matters as much as model performance.

Nice to have:

• Experience hosting and serving your own models.

• A background in NLP, machine translation, OCR, or multilingual or vision-language models.

• Patents or publications at top-tier peer-reviewed conferences or journals.

• Depth in algorithms and data structures, parsing, numerical optimization, data mining, or parallel, distributed or high-performance computing.

• Unix/Linux and professional software development experience.

Eligibility & Location

• Location: on-site in Seattle, WA, 5 days a week. Candidates willing to relocate are welcome, at their own expense. You'd be expected on-site within two weeks of an agreed offer.

• Former Amazon employees are welcome to apply.

• Level: mid-to-senior (equivalent to Amazon L5).

Why This Role

• Ground-floor work on a new science team, shaping how Amazon uses GenAI to reach customers in their own language.

• Your models run in production at huge scale: millions of pieces of content a day across 130+ locales.

• Problems that no off-the-shelf model solves, with the data, compute and support of one of the world's largest ML organizations.