Speech QA Engineer; Vendor Hotword, ASR, TTS, Translate

Manual and semi-automated QA for Android Speech technologies (Hotword / Speaker-, on-device and server ASR, TTS, speech translation) as they are integrated into client surfaces such as Live Caption, Voice Access, Recorder, Gboard voice typing, Live Translate, and Gemini/Assistant. The engineer owns test planning, execution, and bug reporting for speech model and APK releases, and is the first line of reproduction and triage between Speech engineering and client teams.

Test planning

• Write and maintain test plans and test cases from PRDs, design docs, and eng input; keep a reusable test-case library per technology and per client.

• Define coverage matrices: simple (en-US, single account) → medium (additional locales) → complex (multi-language, multi-user, linked devices, accessibility services on).

Release and build verification

• Weekly sanity/regression on latest builds of speech-owned APKs (Speech Recognition & Synthesis / GoogleTTS, Android System Intelligence, Google app) and client apps.

• Pre-release verification of model drops: ASR language packs (SODA), hotword/DSP models, TTS voices, translate models. Confirm model/pack version on device via adb and logcat before testing.

• Pre-dogfood and pre-launch sign-off (P0), in-development testing (P1), post-launch production checks (P2).

Per-technology coverage

• Hotword / Speaker- accuracy, false accepts (e.g. media/adverts), Voice Match enrollment and consent flows, sensitivity settings, battery-saver/offline/driving/locked states, OEM (Samsung) and 1P devices.

• ASR: streaming and dictation quality checks, language switching, on-device vs server fallback, offline behavior, proper nouns/contact names, punctuation, i18n locales.

• TTS: voice selection and download, pronunciation/prosody spot checks, interruption and barge-in, playback latency, accessibility (Talk Back/Select-to-Speak).

• Translate: speech-to-speech and speech-to-text translation in Live Translate and caption surfaces; language pair coverage; latency and rendering.

Client integration testing

• End-to-end CUJs on client surfaces:
  Live Caption (media, calls, captions styling), Voice Access (command recognition, dictation), Recorder (transcription, speaker labels, search), Gboard voice typing, Gemini/Assistant.

• Verify phenotype flag / experiment configurations, and that client apps pick up the expected model and speech APK versions.

• Exploratory testing across device tiers (Pixel generations, tablets, OEM devices), Android versions, and account types (consumer, Dasher/enterprise, supervised/Family Link).

Stress and reliability

• Run long-duration audio playback stress tests (hotword/ASR loops) in the speech lab; report iteration-level failures (DSP errors, OOMs, UI not rendered, mic busy).

• Monitor power/CPU/memory where applicable using Perfetto traces with a documented device state.

Bug reporting and triage

• File actionable Buganizer bugs with repro steps, build/model versions, device, locale, bugreport/logcat/audio artifacts; follow triage routing rules.

• Reproduce and verify eng fixes; keep a QA hotlist and a weekly status report (tested, passed, blocked, open bugs by severity).

Knowledge and enablement

• Maintain testing playbooks (setup, side loading language packs/APKs, debugging tips, known blockers).

• Train other QA teams and client-team testers on speech-specific test methods.

• 3+ years of manual/exploratory QA on Android (mobile or embedded devices).

• Comfortable with adb, logcat, bug reports, flashing/side loading builds and APKs, and toggling developer settings.

• Experience developing test cases and filing high-quality bugs in an issue tracker.

• Able to judge audio/speech quality (clear English; additional languages a plus for i18n spot checks).

Preferred

• Prior QA of voice assistants, ASR, TTS, or accessibility features (Talk Back, Voice Access, Live Caption).

• Basic scripting (Python/shell) to automate repetitive device setup and checks.