ML Engineer

Apple

1d

About The Position

Apple’s Software Localization & Release Engineering team is seeking an ML Engineer to apply large language models and machine learning to software testing and release processes for international users. The engineer will own the model-facing side of AI-powered quality systems by designing LLM-based solutions, evaluating their accuracy, and improving their reliability, efficiency, and trustworthiness. The work supports a fast-moving effort used by internal users and involves collaboration with automation, quality, and localization engineers, as well as international quality engineers, translators, and project managers across several time zones.

Requirements

• Experience building production applications on top of LLMs, including tool use, structured output, or agentic designs.
• Hands-on experience with prompt engineering and systematic model evaluation, such as writing rubrics or graders and measuring precision and false-positive rates.
• Strong programming skills in Swift and/or Python, and experience building backends for ML applications.
• Ability to improve model behavior iteratively based on measured results.
• Hands-on experience fine-tuning or training models.

Nice To Haves

• Experience with computer vision or object detection, such as detecting elements in screenshots or documents.
• Experience with open-weights models and local inference.
• Experience with multi-agent systems.
• Familiarity with LLM serving infrastructure, such as model gateways, retries, and cost-aware model selection.
• Experience building feedback loops that turn model failures into improved prompts, classifiers, or training data.
• Familiarity with software test automation concepts, such as test execution, pass/fail grading, and CI/CD.
• Experience in software localization or internationalization.
• Experience mentoring other engineers.

Responsibilities

• Build evaluation frameworks, including benchmarks, ground-truth datasets, and graders; measure and improve precision, recall, and false-positive rates.
• Reduce the cost and latency of each run through prompt improvements, system design, and appropriate model selection for each stage.
• Build feedback loops that turn confirmed errors into improved prompts, classifiers, or training data.
• Design and build LLM-based systems that perform quality tasks and independently grade their results, including adversarial grading approaches that catch false-pass results.
• Design classification pipelines to triage failures and distinguish product issues from test or tooling problems.
• Build datasets for evaluating, training, or fine-tuning models.
• Write clean, reviewable code and help maintain a fast-moving codebase as it grows.

What This Job Offers

• Job Type: Full-time
• Education Level: No Education Listed