Voice Pipeline
300 ms latency
Stakeholder Reviews
20+
Team Size
3 engineers
System ArchitectureINTERACTIVE_DIAGRAM
Summary
AI-powered ATC-pilot simulator for Honeywell Aerospace's Anthem ecosystem. Real-time voice pipeline at 300 ms (LiveKit, Deepgram, ElevenLabs) paired with Python voice analytics: YIN pitch, VAD, z-score calibration, and abstention scoring under 0.60. CBTA dashboard surfaces cognitive load and communication competency to instructors.
voice_analytics.pymic
frame = audio.next()
pitch = yin(frame)
score = z_score(
pitch, baseline
)
if abstention(score) < 0.6:
grade = cbta(transcript)
dashboard.push(grade)videocamProject Demo
VIDEOMain Demo
rocket_launchFuture Prototypes
boltFuture Prototypes Demoopen_in_newAdaptive AI
Pilot Predict
Scenario Builder
Layout Customisation