Velvet
Training data for underserved domains
Y CombinatorFall 2025ActiveB2BArtificial IntelligenceGenerative AIData EngineeringConversational AI
What they do
Problem
AI teams in underserved domains lack sufficient high-quality training data to build effective models.
Solution
Velvet provides training data tailored to underserved domains for machine learning model development.
Who it's for
Machine learning teams and organizations building AI models in underserved domains.
Founders
Lucas Mantovani
Founder
Co-founder, CEO of Velvet. Ex video data + youngest member @ Meta FAIR. CS honors @ UChicago. Brazilian olympiad medalist in Math + Physics.
Lucas Tucker
Founder
Co-founder, CTO of Velvet. Ex-eng @ Adobe working on realtime content delivery. Honors Math & CS @ UChicago.
Open roles 3
Company timeline
- Founded2025
- YC batchFall 2025
- StatusActive
- Team size3 people
- LocationSan Francisco, CA, USA
Similar startups
Mundo AI
High Quality Multilingual Training Data for AI Models
Shofo
The World's Largest Video Library
Sureform
Real-world data for multimodal and embodied AI
Poseidon
We’re leading a $15M seed round in Poseidon, which is building a dece…
Datacurve
Frontier coding data for training and evaluating LLMs
Datoric
Responsibly sourced data for multimodal AI