Velvet is focused on building conversational models that are engaging.
Founders
- Founded
- 2025
- Team
- 3 people
Browser storage is unavailable. Keep this page open and export your collection before leaving.
Velvet is focused on building conversational models that are engaging.
Founders
We're a research lab building the intelligence that is intrinsically motivated to learn and share, without human tutors.
Product
Mantis builds the infrastructure layer for living, predictive models of individual human beings — forecast to the horizon where a decision still has to be made. Starting with medicine.
Founders · Product
An applied AI research lab building dependable, domain-grounded agents.
Founders · Product
Aviro designs challenging worlds for AI models to explore and solve problems, delivered as post-training datasets and reinforcement learning environments.
Product
Natural voices in 39 languages, developed and hosted in Europe, fully GDPR compliant.
Founders
Panels provides high-quality audio datasets for training and evaluating speech and audio models. We work with frontier labs and early stage startups building at the edge of Voice AI.
Customers
Brainbase lets teams get to extremely powerful agents extremely fast across any model or harness by handling agent hosting, scaling, evaluations and more.
Founders · Pricing
Petrarch helps operators across the critical infrastructure chain prepare their proprietary operational records, cleared and anonymized on premise, so the AI systems building for the physical world can learn from real work.
Customers · Founders · Product
Compare computer-use AI agents by blind preference, completion, speed, recovery and cost. Explore live model rankings or post a task and judge a battle.
Founders · Product
Human Archive (YC W26) is building the data and deployments infrastructure for physical AI. We design hardware, deploy cameras and sensors globally, and train models that help AI better understand how humans interact with the…
Founders · Product
Custom conversational audio datasets, expert annotation, and licensed voice actor data for training and evaluating speech and voice AI models.
Founders · Product · How it works
MangoDesk — Accelerate custom evals and post-training data creation.
Founders
Mundo develops data, evaluations, and applied research for frontier labs and AI companies to advance perceptual intelligence across audio, video, and emerging modalities.
Founders
Markov is the data lab for engineering and computer-use. We capture real work from domain experts and craft high-fidelity datasets and RL environments.
Founders · Product
Sureform helps companies transform their internal operational data into high-quality RL environments and de-identified datasets that can be licensed to leading AI labs for model training.
How it works
We're building a protocol for large language models to directly interface with image sensors.
Custom data for long-horizon reasoning, software engineering, and data science.
Founders
Halluminate builds RL environments for financial services — computer-use and tool-use gyms that train frontier AI models on real investment banking, private equity, and consulting work.
Founders · Product
Luel is a multi-modal lab building the foundational models of machine perception.
Product
Environments and data for the next generation of AI. Klavis AI builds RL environments and high-quality training data at scale across long-horizon coding, AI4Science, RSI, knowledge work, and agentic tool use.
Founders
CTGT is an applied AI research laboratory fundamentally solving the alignment and reliability bottleneck for enterprise AI. For enterprises, especially highly regulated industries, deploying Generative AI is historically a compromise…
Founders · Product · How it works