Liva AI
Audio & Video Data
Y CombinatorSummer 2025ActiveB2BArtificial IntelligenceMarketplaceB2BData Labeling
What they do
Problem
Speech models trained on broad internet data lack realistic, human-sounding performance due to insufficient targeted training data.
Solution
The company collects targeted audio and video speech datasets for AI model labs to train and improve speech models.
Who it's for
AI model labs and companies developing speech and multimodal models.
Founders
Ashley Mo
Founder
Co-founder & CEO | Caltech CS dropout | Published research on an AI model that could detect lung disease through cough sounds at 17 y/o (IEEE Xplore) and led research study to collect hard-to-get patient audio data @MIT at 18 y/o | Built biosensors for chronic kidney disease @Caltech and another for iron deficiency (for fun!). I also like dance!
Aoi Otani
Founder
Co-founder & CTO | Harvard Computer Science & Biology
Open roles 3
Company timeline
- Founded2025
- YC batchSummer 2025
- StatusActive
- Team size2 people
- LocationSan Francisco, CA, USA
Similar startups
floatingpoint
Post-training data to teach models document work
Nitrode
Data for reasoning agents in dynamic environments
PerfectBit
Training data for frontier AI labs
Specific
Audio and Enterprise Human Data for AI
Sciloop
Expert STEM reasoning data for frontier AI labs.
hillclimb
Training Data for Recursive Self-Improvement