Want to know if this job is worth applying to?
San Francisco, California, United States
Onsite
Available
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.
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.
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'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.
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.
We strongly encourage you to apply even if you don't meet every qualification. The strongest candidates we meet rarely do, so don't exclude yourself prematurely. We're rethinking how humans interact with AI, and doing that well demands diversity of perspective and experience.
We're committed to a fair and accessible interview process. If you need any accommodations or adjustments, please let us know.
