Lead Engineer, AI Platform

United States of AmericaRemoteLead

The AI Quality engineering team at Circle owns the foundation for measuring, diagnosing, and improving the quality of Circle's AI-powered features. This team focuses on building the infrastructure to measure, diagnose, and improve production AI systems, rather than ML research or model training.

We're looking for a Lead Engineer to help us build out the evaluation frameworks, observability tooling, and diagnostic infrastructure that tell us whether our AI Agents are working well, where to improve them, and how to make them faster and more cost-efficient.

This is a hands-on player-coach role. You’ll also lead and manage the AI Quality engineering team, with an ambitious and growing roadmap.

If you're excited about making AI systems work reliably, efficiently, and at scale, this is for you.

What you'll be doing

Build and own our evaluation infrastructure. Design the CI/CD pipelines, scorers, and datasets that tell us whether Circle's AI agents (planners, tool-callers, and sub-agents) are actually working, from a single tool call to a full multi-turn conversation.
Diagnose exactly where quality breaks down. Trace failures across the agent pipeline including plan creation vs. execution, tool selection, tool trajectory in complex areas like workflows, site builder, and analytics. Turn what you find into prototypes that solve the issue or identify priorities for AI core engineering can help.
Grow the datasets that make evaluation possible. Stand up annotation workflows and build out AI-generated and simulated conversations so we can cover more of the product faster than manual labeling alone.
Run structured experiments across prompts, models, and the agent harness. Evaluate new and open-source models against our production baseline, build the framework we use to decide when to shift models, and chase cost and latency wins through model swaps, caching, and routing by plan complexity.
Lead and grow the AI Quality engineering team. Set technical direction and manage day-to-day priorities, all while staying hands-on in the code yourself.
Partner closely with AI Core engineering. Work with the engineers building Circle's AI products so they have real confidence that their changes are actually improving quality, not just shipping.

What you'll need to be successful

7+ years of experience building and shipping production software, ideally including LLM-powered agents that take real actions in a product. You've worked on complex, tool-using systems (multiple tools, planning or orchestration, sub-agents) not just simple, single-turn assistants, and you can walk us through something you shipped and how you knew it was actually working.
Comfortable in Ruby on Rails / Python or ready to pick them up quickly. Ruby on Rails is our production system and the foundation that Circle’s AI Agents are built on and proficiency in Python is a strong plus, especially for the data and evaluation side of the work.
Experience building evaluation or observability infrastructure for ML/AI systems. You've built eval pipelines, scorers, dashboards, or CI/CD for evals before and have experience with evaluation frameworks like Braintrust, LangSmith, or similar.
Experience designing datasets, annotation workflows, or labeling pipelines for ML/AI evaluation. You know how to turn raw examples into a dataset you can trust through human annotation, AI-generated data, or simulation.
Learn fast, experiment aggressively, and thrive in a highly dynamic environment. You’ll need to be comfortable running a lot of experiments and letting the results settle the argument, including the ones that don't work (which will be many of them in the beginning).
Comfortable in a fast-paced environment with ambiguity. You’ll need to learn and pick up new technologies when projects require it.
Strong alignment with our company values.
You are proficient in English (spoken, written, and reading) at a CEFR Level C2 / ILR Level 5.

What you'll be doing

Build and own our evaluation infrastructure. Design the CI/CD pipelines, scorers, and datasets that tell us whether Circle's AI agents (planners, tool-callers, and sub-agents) are actually working, from a single tool call to a full multi-turn conversation.

Diagnose exactly where quality breaks down. Trace failures across the agent pipeline including plan creation vs. execution, tool selection, tool trajectory in complex areas like workflows, site builder, and analytics. Turn what you find into prototypes that solve the issue or identify priorities for AI core engineering can help.

Grow the datasets that make evaluation possible. Stand up annotation workflows and build out AI-generated and simulated conversations so we can cover more of the product faster than manual labeling alone.

Run structured experiments across prompts, models, and the agent harness. Evaluate new and open-source models against our production baseline, build the framework we use to decide when to shift models, and chase cost and latency wins through model swaps, caching, and routing by plan complexity.

Lead and grow the AI Quality engineering team. Set technical direction and manage day-to-day priorities, all while staying hands-on in the code yourself.

Partner closely with AI Core engineering. Work with the engineers building Circle's AI products so they have real confidence that their changes are actually improving quality, not just shipping.

What you'll need to be successful

7+ years of experience building and shipping production software, ideally including LLM-powered agents that take real actions in a product. You've worked on complex, tool-using systems (multiple tools, planning or orchestration, sub-agents) not just simple, single-turn assistants, and you can walk us through something you shipped and how you knew it was actually working.

Comfortable in Ruby on Rails / Python or ready to pick them up quickly. Ruby on Rails is our production system and the foundation that Circle’s AI Agents are built on and proficiency in Python is a strong plus, especially for the data and evaluation side of the work.

Experience building evaluation or observability infrastructure for ML/AI systems. You've built eval pipelines, scorers, dashboards, or CI/CD for evals before and have experience with evaluation frameworks like Braintrust, LangSmith, or similar.

Experience designing datasets, annotation workflows, or labeling pipelines for ML/AI evaluation. You know how to turn raw examples into a dataset you can trust through human annotation, AI-generated data, or simulation.

Learn fast, experiment aggressively, and thrive in a highly dynamic environment. You’ll need to be comfortable running a lot of experiments and letting the results settle the argument, including the ones that don't work (which will be many of them in the beginning).

Comfortable in a fast-paced environment with ambiguity. You’ll need to learn and pick up new technologies when projects require it.

Strong alignment with our company values.

You are proficient in English (spoken, written, and reading) at a CEFR Level C2 / ILR Level 5.