The problem you own

The system cannot decide lawfulness before processing the audio, and it is forbidden from durably recording the audio before it decides. Everything you build lives inside that contradiction. You own the audio spine: capture connectors for Meet, Zoom, Teams, SIP legs, and desktop audio; streaming transcription on models we host and tune ourselves; and a latency budget measured against the buffer horizon, perception latency plus grace plus the pre-roll of every armed trigger.

What you will do

Design and own the real-time capture and streaming pipeline end to end, over WebRTC and WebSockets, including telephony legs with no per-participant stream to lean on.

Make frame-size, chunking, and endpointing decisions from evidence, and defend them in design review.

Hold accuracy across accents, crosstalk, and terrible conference-room microphones.

Co-design the interface to the gate so classification runs inside the transient buffer window, never after it.

What great looks like

You have shipped production streaming speech at scale, you argue about frame sizes from experience, and p95 latency is a number you track without being asked.

Stack and signals

Python plus a systems language (C++, Rust, or Go), CUDA a plus, WebRTC, gRPC or WebSockets, self-hosted ASR.

Compensation

Cash is deliberately founder-grade; the equity is the point, and it is real. You will work directly with the founding technical leadership, in the codebase, from week one.

How we hire

One conversation about your best work, one deep-dive on a system or a deal you owned, one paid working session on a real Cull problem, one founder conversation. Usually under two weeks.