SDE II

Responsibilities:

• A real feature inside the assistant, shipped to real users. You'll have design help, but the feature will be yours.
• Getting it into production on our weekly deploy cadence, in front of people using it in their own language on a real network.
• The instrumentation that tells you whether it worked and a clear answer when somebody asks how it is doing.
• Real user data, handled with the care an identity company demands.
• We're AI-first in how we build. Even at this level, we expect you to design the workflow rather than simply run the tool.
• You build workflows and skills that your teammates reuse, so a good result becomes repeatable.
• You treat context as a budget you spend deliberately, and you can explain what you put in front of a model and what you deliberately left out.
• You're curious about taking on work that used to be too large for one person, and you understand that what makes such a change safe is the verification you build first, not the speed of the generation.
• You keep up with what has recently become possible, and you form your own view.

Requirements:

• Around 4-8 years of building software. We care about what you've actually built.
• Something you took from nothing to real users. That it existed and somebody used it matters far more than whether it succeeded.
• Strong fundamentals in code, data structures, and debugging.
• The habit of finishing. We're looking for the ones who took it all the way to somebody using it.
• Taste, and the honesty that goes with it. You should notice when something feels wrong and say so, even when it is your own work.
• Our stack: Python (core services), Rust (latency-critical media path), AI harness (agent loop, tools, prompts, evals), voice streaming (persistent websocket sessions), model providers (speech and language), AWS, Postgres, Redis, and Kotlin/Swift (apps). You don't need to have used any of it; how you think matters most.