Job description

Momentic is building the definitive verification layer for software: an AI-native testing platform that helps engineering teams catch bugs before they ship. As AI accelerates software development, teams need a reliable way to verify that their products work. Engineering teams at Notion, Webflow, Xero, and Whatnot use Momentic to ship with confidence. You’ll join our engineering team and work closely with our CTO to improve the models and agents behind Momentic. We’re looking for a hands-on AI engineer who can fine-tune models, build evaluations, and turn production data into more reliable, efficient agents. Momentic executes more than 1 million tests daily and processes more than 1 billion tokens per hour at peak, giving your improvements a direct impact on the product. WHAT YOU’LL DO - Fine-tune models for long-running agents, including agents that investigate whether test failures come from infrastructure, recent code changes, or test authorship. - Build and curate training and evaluation datasets from production workflows, using failure patterns to identify opportunities for improvement. - Develop LLM evaluations and observability that connect offline results with production behavior and help the team verify improvements. - Tune prompts, context, and model behavior to improve task success, reliability, latency, token usage, and cost. - Design repeatable experiments and help publish benchmarks that demonstrate our core agents’ performance on testing and browser workflows. - Own technical projects from problem definition through delivery, working with the engineering team to make practical tradeoffs and ship improvements. YOU’RE PERFECT FOR THIS ROLE IF… - You have relevant software engineering experience, typically 3+ years, including hands-on contributions to AI-powered capabilities used in production. - You have hands-on experience with supervised fine-tuning, including preparing training data and evaluating model behavior. - You’ve worked with prompt and context tuning, evaluation datasets, and LLM observability to diagnose failures and measure improvements. - You have strong engineering fundamentals, code efficiently, and know what you’re great at and where you have room to grow. - You use AI agents to build software, breaking down work, tracking progress, checking results, and redirecting agents when they get stuck. - You’re comfortable with autonomy and ambiguity. You can turn an unclear problem into a plan, manage your own work, and follow through to delivery. We’re interested in the depth of your contributions and what you’ve shipped. Your experience may come from owning an AI capability within a larger product. NICE TO HAVES - Experience with reinforcement learning fine-tuning, RL environments, or LLM-as-judge evaluators. - Experience with long-running agents, recovery from failed steps, or coordinating multiple agents. - Experience publishing benchmarks or running rigorous experiments on AI systems. ABOUT US - Raised a $15M Series A led by Standard Capital, with backing from Y Combinator, FCVC, Transpose Platform, and Dropbox Ventures. - Our customer base has grown 4× year to date. - A team of 17 and growing. - Based in person in San Francisco at 650 5th Street. BENEFITS - Medical, dental, and vision insurance. - 401(k). - Unlimited PTO and sick days, plus company-observed US holidays. - Daily lunch and a fully stocked kitchen. - Commuter benefits. - Team offsites and events.

Requirements

Must-haves

Fine-tune models for long-running agents, including agents that investigate whether test failures come from infrastructure, recent code changes, or test authorship.
Build and curate training and evaluation datasets from production workflows, using failure patterns to identify opportunities for improvement.
Develop LLM evaluations and observability that connect offline results with production behavior and help the team verify improvements.
Tune prompts, context, and model behavior to improve task success, reliability, latency, token usage, and cost.
Design repeatable experiments and help publish benchmarks that demonstrate our core agents’ performance on testing and browser workflows.
Own technical projects from problem definition through delivery, working with the engineering team to make practical tradeoffs and ship improvements.
You have relevant software engineering experience, typically 3+ years, including hands-on contributions to AI-powered capabilities used in production.
You have hands-on experience with supervised fine-tuning, including preparing training data and evaluating model behavior.
You’ve worked with prompt and context tuning, evaluation datasets, and LLM observability to diagnose failures and measure improvements.
You have strong engineering fundamentals, code efficiently, and know what you’re great at and where you have room to grow.
You use AI agents to build software, breaking down work, tracking progress, checking results, and redirecting agents when they get stuck.
You’re comfortable with autonomy and ambiguity. You can turn an unclear problem into a plan, manage your own work, and follow through to delivery.
We’re interested in the depth of your contributions and what you’ve shipped. Your experience may come from owning an AI capability within a larger product.

Nice-to-haves

Experience with reinforcement learning fine-tuning, RL environments, or LLM-as-judge evaluators.
Experience with long-running agents, recovery from failed steps, or coordinating multiple agents.
Experience publishing benchmarks or running rigorous experiments on AI systems.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

aillm

Platform: ashby
Effort: quick

Benefits

Benefits detected in the listing.

Dental

Vision

401(K)

Pto