AI Engineer, Internal Systems
First engineer building the verification and evaluation layer for Wispr's internal AI coding agents — creating end-to-end verification loops, fast test environments, eval signals from failures and human feedback, and operating the agent fleet as a production system.
- Compensation
- $220K–$300K / yr
- Location
- San Francisco
- Work setup
- Onsite
- Seniority
- Not specified
- Employment
- Not specified
- Experience
- Not specified
- Category
- Ai engineering
- Posted
Skills
Must have
- Built internal tools that other engineers adopted and can explain how their impact was measured
- Uses agentic coding tools deeply with specific opinions about where they succeed and fail
- Has built evaluation or verification infrastructure such as test harnesses, CI systems, eval pipelines, or benchmarks
- Can work hands-on across application code, infrastructure, and unfamiliar systems
- Empirical about own work, attentive to subtle failure modes, comfortable owning ambiguous problems
Nice to have
- Prior experience in voice, model training, or the company's product domains is not required
Full description
About Wispr
Wispr is an AI research and product company building the voice interface for computing. AI can now reason, code, and act. Yet, humans still do the work of the interface. We think that’s backwards.
Our first products are Flow, which lets you speak naturally in any application, and Notetaker, which builds context across conversations. We’re building toward an interface that can perceive, understand, and take action with earned trust. That means solving hard problems across models, systems, and product - and caring about the final human experience as deeply as the technology underneath it.
We’re a talent-dense team that holds strong opinions, tests them quickly, and builds technology that sparks joy. Our goal is to build the first voice interface used every day by a billion people.
About the role
Our agents already write code and respond to incidents around the clock, but they can't yet tell when they're done, so every loop still ends with a human checking. You'd be the first to build the verification and evaluation layer that lets them finish with confidence.
What you'll do
- Build end-to-end verification loops that let agents determine when their work is actually complete
- Create fast, reliable testing environments and turn failures and human feedback into durable evaluation signals
- Improve how context, instructions, skills, and memory are delivered to agents—and measure what makes them perform better
- Operate the agent fleet as a production system and measure its quality, adoption, and impact
You may be a fit if
- You've built internal tools that other engineers adopted, and you can explain how you measured their impact
- You use agentic coding tools deeply and have specific opinions about where they succeed and fail
- You've built evaluation or verification infrastructure such as test harnesses, CI systems, eval pipelines, or benchmarks
- You can work hands-on across application code, infrastructure, and unfamiliar systems
- You are empirical about your own work, attentive to subtle failure modes, and comfortable owning ambiguous problems
Prior experience in voice, model training, or our product domains is not required.
Get to know us
- Our operating principles
- Our Series B
- Advancing HCI
- Introducing Wispr Advanced Interfaces Lab
Logistics
We sponsor H1B, O1, EB1, L1, STEM OPT, and more. We can't guarantee sponsorship for every role, but if we make you an offer we'll make every reasonable effort, with help from our immigration law firm.