Senior AI Platform Engineer - Scale LLM & Agent Workflows

Snorkel AI; San Francisco (CA); On-site; USD 210,000 - 320,000; Full time

Job summary

Snorkel AI is building the platform that enables teams to run large-scale AI experiments and production workloads. You will design and operate distributed systems for agentic workflows, including synthetic data pipelines, evaluation infrastructure, and multi-model orchestration at scale.

The role requires 5+ years in production software, strong Python skills, and a background in distributed systems and cloud platforms.

Qualifications

• 5+ years building production software systems, with experience in AI/ML infrastructure, ML platforms, distributed systems, data platforms, or backend infrastructure
• Experience operating non-deterministic AI or ML workloads in production or at significant scale — you are comfortable reasoning about behavior across models, tools, environments, and multi-step execution
• Experience building infrastructure for experimentation, evaluation, model development, synthetic data, agentic workflows, training, inference, or production ML systems
• Strong proficiency in Python and experience building production-quality APIs, services, and developer tooling
• Strong background in distributed systems and cloud platforms (AWS preferred), including compute orchestration, storage, networking, isolation, and failure handling
• Experience with workflow or distributed execution frameworks such as Prefect, Airflow, Dagster, Ray, Kubernetes, or similar systems
• Strong understanding of production system fundamentals - observability, telemetry, reliability, performance, debugging, incident response, and cost management
• Ability to reason about AI system quality beyond traditional service metrics, including evaluation design, experiment reproducibility, behavioral regressions, and model or agent variability
• Track record of leading complex engineering initiatives, influencing stakeholders, and delivering measurable impact
• Ability to work in a fast-paced environment with strong technical communication skills
• Fluency with modern AI and developer tooling and a willingness to rapidly evaluate and adopt new models, frameworks, infrastructure, and techniques as the ecosystem evolves

Responsibilities

• Design and build infrastructure for running large-scale agentic workloads, including multi-step agents interacting with tools, external services, sandboxes, and simulated environments
• Build scalable synthetic data generation and automated labeling systems that allow teams to create, refine, and evaluate high-quality training and evaluation datasets
• Design evaluation infrastructure for measuring AI system behavior across models, prompts, tools, environments, and multi-step trajectories - including reproducible experiments, benchmark execution, regression detection, and continuous evaluation
• Build orchestration and distributed compute systems for running thousands to millions of AI experiments and simulations reliably across heterogeneous compute environments
• Develop infrastructure for agent simulation environments, including environment provisioning, isolation, lifecycle management, and scalable execution
• Build and operate LLM infrastructure for routing, rate limiting, retries, caching, provider failover, cost attribution, and efficient execution across multiple model providers
• Instrument agent and model workloads so failures are observable and debuggable - capturing traces, model interactions, tool calls, environment state, evaluation results, latency, reliability, and cost
• Design systems that make non-deterministic workloads reproducible and measurable, allowing engineers to compare experiments, diagnose behavioral regressions, and understand why an agent succeeded or failed
• Improve the developer experience for AI experimentation by building APIs, SDKs, workflow abstractions, and tooling that make it easy to move workloads from local development to large-scale production execution
• Collaborate with research, product, and engineering teams to turn experimental AI workflows into reliable, reusable platform capabilities

Skills

Production software

Non-deterministic AI

Experimentation infra

Python

Distributed systems

Cloud platforms

Workflow frameworks

Observability

AI system quality

Leadership

Tech communication

Tools

Prefect

Airflow

Dagster

Ray

Kubernetes

AWS

Job description

Snorkel AI is building the platform that enables teams to run large-scale AI experiments and production workloads. You will design and operate distributed systems for agentic workflows, including synthetic data pipelines, evaluation infrastructure, and multi-model orchestration at scale.

The role requires 5+ years in production software, strong Python skills, and a background in distributed systems and cloud platforms.