The technical challenge:

The interesting problems aren't in the prompt anymore. They're in the system around it: what the agents are allowed to do, how the pieces fit together, where the guardrails go, and what 'correct' even means when the output isn't deterministic.

We are building the first AI-powered advisor to manage all your physical assets. We started with the car in 2019, now we're expanding into new verticals (e.g. home, motorcycle, RV, boat) and accelerating our growth.

Before off-the-shelf platforms existed, we built our own infrastructure to automate sales and service conversations. That system now handles over 70% of inbound sales and service requests: more than 50,000 chats a month, in production, at scale.

That head start gave us real leverage, and it's outgrown ad hoc ownership. We need someone who can run a fast-moving, high-volume chatbot operation day to day and has a real point of view on the next bets: voicebot, computer use, consumer AI applications.

What you'll actually do:

• Decide what we build next in AI, and make the case for it. Nobody hands you a roadmap.
• Partner with OpenAI to evaluate and deploy next-generation voice models and workflow automation.
• Evaluate emerging AI technology (voicebot, computer use agents, and beyond), prototype what's worth building, and make the call on what we invest in next.
• Design evaluation frameworks and guardrails, and make the real tradeoffs between latency, cost, and accuracy in a regulated production environment.
• Set the technical bar for our chatbot systems - shape what the prompt engineering team builds next and keep the architecture sound as it scales.
• Build new automations across areas like record disputes, email/callout reduction, document validation (ex. licenses, proof of insurance).
• Work directly with subject matter experts to turn domain knowledge into systems that behave correctly.

Who you are:

• A systems designer. You think in interfaces, failure modes, and blast radius. You've been the person who redesigned something messy rather than working around it.
• Fluent in modern AI. You read the release notes when a new model drops and you've already built something with it. You have opinions about eval design, and they came from doing it.
• Pragmatic. You know when to build the durable thing and when to ship the version that works this quarter.

Ideal experience:

• 5+ years in a product engineering or forward deployed engineering, ideally in an early stage startup
• A track record of turning ambiguous problems into systems that actually run
• Hands-on experience putting LLM-based systems into production, including the unglamorous parts: evals, regressions, latency, cost, and failure handling

While we appreciate your interest and application, due to the volume of applications we receive, only applicants under consideration will be contacted.

Please note: Jerry.ai is only able to hire full-time salaried employees who currently reside in AZ, CA, CO, FL, GA, IL, MA, NC, NJ, NV, NY, OR, TN, TX, UT, VA, or Ontario, Canada, and full-time hourly employees who currently reside in AZ, FL, GA, NV, NY, NC, TN, TX, UT, or VA. Unfortunately, we cannot accept applications from candidates residing elsewhere at this time.

The successful candidate's starting pay will fall within the pay range listed on this job posting, determined based on job-related factors including, but not limited to, skills, experience, qualifications, work location, and market conditions. Ranges are market-dependent and may be modified in the future. In addition to base salary, the compensation may include opportunities for equity grants.

We offer a comprehensive benefits package to regular employees, including health, dental, and vision coverage, paid time off, paid parental leave, 401(K) plan with employer matching, and wellness benefits, among others. Equity opportunities may also be part of your total rewards package. Part-time, contract, or freelance roles may not be eligible for certain benefits.