Skip to main content
Apple

AIML - Senior Machine Learning Platform Engineer, Responsible AI and Safety

Apple Cupertino, California, United States 2 days ago
Join us in shaping the future of Generative AI at Apple! Are you passionate about making AI systems safer, more inclusive, and globally representative? Apple is seeking a Machine Learning Platform Engineer to build the tooling and infrastructure behind our Responsible AI mitigations, from on-device foundation models running on Apple Silicon to server-side inference on Private Cloud Compute (PCC). In this role, you will build the Python platform our ML and safety engineers rely on to train, evaluate, deploy, and configure safety classifiers, guardrail models, and model behavior overrides. You will turn manual, one-off workflows into automated, scalable, and reproducible systems, so that every evaluation runs the same way, every deployment is validated, and every configuration change is safe to ship

Our team leads Responsible AI initiatives for global generative AI products, operating at the intersection of policy, product, and GenAI. We build the safety guardrails, and policy enforcement layers that protect users from unintended model behavior. This role builds the platform underneath that work. You will build the systems that enable consistent training and evaluation at scale: an training/evaluation platform that runs any benchmark the same way every time, deployment pipelines that package and release models and mitigation assets to devices and the cloud, and tooling that lets engineers across many languages and modalities author and validate configuration changes with confidence. You will also dig into the failures and regressions the platform surfaces, and turn them into concrete fixes. ML engineers, client engineers, policy, and release teams are your customers, and you will need enough understanding of how models are evaluated and served to build infrastructure whose results they can trust.

Evaluation Platform: Design and build a scalable, automated platform for running evaluations of safety classifiers, guardrail models, and overrides consistently, including job orchestration, reproducible run configurations, versioned datasets and results, and regression reporting between model and configuration versions\\nModel Deployment Infrastructure: Own the infrastructure responsible for packaging, versioning, validating, and shipping production ML models and mitigation assets across release trains, from build through rollout and rollback\\nConfiguration \u0026 Overrides Tooling: Build the schemas, validation, and automated tests for the configuration files that control model behavior (overrides, deny lists, thresholds, per-feature policy), so every change is checked for correctness, scope, and unintended impact before it reaches users\\nInference Tooling: Build Python libraries, services, and CLIs that let teams run guardrail and foundation models the same way production does, including preprocessing and normalization parity, so evaluation results represent what ships\\nError Analysis \u0026 Data Pipelines: Build the pipelines that manage evaluation datasets and results, run error analyses on the failures and regressions they surface, and turn findings into concrete model, data, or configuration fixes, while ensuring data handling meets Apple privacy and retention policies\\nPlatform Adoption \u0026 Automation: Make the platform self-serve for adopting teams, automating repetitive authoring and review steps (including with LLM-assisted and agentic tooling) so engineers spend their time on judgment, not mechanics\\nCross-Functional Collaboration: Partner with ML, client, and release engineering teams to understand their workflows and translate them into reliable, performant infrastructure, and represent the platform in architecture and design reviews

3+ years of professional engineering experience, with Python as a primary development language\\nBS in Computer Science, Mathematics, Statistics, or a related field, or equivalent industry experience\\nExperience building infrastructure or tooling for ML inference, evaluation, or deployment that is used by engineers outside your direct team\\nExperience building or operating infrastructure that deploys ML models to production and maintains them across versions\\nWorking knowledge of how ML models are evaluated (metrics, held-out test sets, error analysis, sources of non-reproducibility), enough to build evaluation infrastructure whose results ML engineers can trust\\nExperience designing configuration-driven systems, including schemas, validation, versioning, and safe rollout of config changes\\nStrong software engineering fundamentals: testing, CI/CD, code review, and writing maintainable code that others build on\\nExcellent communication skills and the ability to explain technical tradeoffs to cross-functional partners\\nMust be comfortable working with sensitive or potentially offensive content

Experience with LLM inference, including tokenization, batching, streaming, and serving frameworks\\nWorking knowledge of on-device ML runtimes (Core ML, MLX, or equivalent) and the model-export lifecycle: converting trained models into shippable assets and loading them at runtime\\nExperience running distributed evaluation or workflow orchestration at scale (Kubernetes, Ray, Spark, or equivalent)\\nExperience with text normalization, pattern-based filtering, or multilingual NLP\\nExperience with Responsible AI, trust and safety, or content moderation systems\\nExperience building LLM-based or agentic developer tooling\\nFamiliarity with Swift
Apply now
Cupertino, California, United States
On-site
2 days ago

Share this job