Brave AI
Modular AI Data Infrastructure

Brave AI Data Collection Campaigns

The platform powers high-integrity AI dataset collection. Currently, active collection is live for the "Hey Brave" wake-word engine. Additional campaigns are in planning and will launch as protocols are decided.

Live Active Collection

Wake-Word Engine: "Hey Brave"

Collect isolated wake words, continuous commands, phonetically near confusables, natural free speech, and ambient room tone.

Section A: 20 Isolated Takes
Section B: 10 Command Takes
Section C: 22 Confusables
Section D: 90s Free Speech
Section E: 30s Room Tone
Modality: Audio (16-bit 16kHz PCM WAV)
Launch Studio
Coming Soon

Conversational Turn-Taking & Audio

Future multi-turn conversational voice data and dialogue latency calibration. Collection protocols and criteria are currently under development.

Campaign protocol in design
Acoustic latency calibration pending
Contribution criteria to be announced
Modality: To Be Announced (TBA)
Coming Soon
Coming Soon

Search & Citation Evaluation

Evaluation benchmarks and human preference feedback for search synthesizers. Campaign scope and rubrics will be announced upon engineering review.

Evaluation methodology in development
Citation verification rubric planning
Testing timeline to be determined
Modality: To Be Announced (TBA)
Coming Soon
Coming Soon

Agentic Systems & Task Benchmarks

Future evaluation benchmarks for agentic browser workflows and multimodal reasoning. Specific criteria will be published once officially decided.

Task scope & evaluation criteria in review
Privacy sandbox integration pending
Pipeline schedule to be announced
Modality: To Be Announced (TBA)
Coming Soon