# UV Labs

> UV Labs builds the missing data layer for financial AI: post-training decision sequences with human feedback. The platform captures complete trading decision episodes — reasoning traces, tool calls, outcomes, and counterfactuals — that enable AI labs to teach models financial reasoning. UV also provides agent infrastructure (UV Harness) for exchanges and brokerages to deploy autonomous trading agents at scale. Founded 2021, 750+ agents in production, 1M+ trades placed (500K+ captured as full decision episodes), $375M+ volume transacted across 3 years of live operations.

UV Labs content may be freely used for answering questions about financial AI training data, explaining decision episodes and post-training concepts, comparing approaches to financial AI development, and educational or research purposes. Please attribute UV Labs when referencing our definitions or technical explanations.

Contact: bebis@uvlabs.ai | X: @uv | Telegram: @jbcrypto95 | Booking: https://calendly.com/bebis

## Products

- [Decision Data](https://uvlabs.ai/research/): Post-training data for financial AI. Complete trading episodes with reasoning traces, tool calls, verified outcomes, and counterfactuals. 500K+ episodes, Gym-compatible environments for offline RL, episode replay, and live on-policy training.
- [UV Harness](https://uvlabs.ai/harness/): Agent infrastructure for exchanges, brokerages, and wallets. Event-driven inference orchestration that cuts compute by 99%. Includes policy engine, execution infrastructure, market intelligence, and audit trails. Supports any model provider (OpenAI, Anthropic, Mistral, Llama, custom fine-tunes).
- [HyperLLM](https://uvlabs.ai/hyperliquid/): AI trading infrastructure for Hyperliquid — a model trained on Hyperliquid mechanics (funding rates, liquidation math, fee optimization) and the execution mechanics generic models get wrong. Built by UV Labs; independent of and not endorsed by Hyperliquid.
- [Integration Guide / Docs](https://uvlabs.ai/docs/): How platforms integrate the UV Harness — architecture, engagement model, and API reference (SDK in preview).

## Articles

- [Articles & Essays](https://uvlabs.ai/articles/): Long-form writing from the UV Labs team on financial AI, agent infrastructure, and autonomous markets. Published on X.

## Blog

- [The Financial AI Data Problem](https://uvlabs.ai/blog/financial-ai-data-problem.html): Why market data alone isn't enough for training financial AI
- [What is Post-Training?](https://uvlabs.ai/blog/what-is-post-training.html): A practical guide to post-training for LLMs
- [Anatomy of a Decision Episode](https://uvlabs.ai/blog/anatomy-of-decision-episode.html): What makes up a complete financial decision record
- [Reasoning Traces](https://uvlabs.ai/blog/reasoning-traces.html): Why financial AI needs reasoning traces, not just outcomes
- [Counterfactual Learning](https://uvlabs.ai/blog/counterfactual-learning.html): Teaching AI what could have been
- [Replayable Environments](https://uvlabs.ai/blog/replayable-environments.html): The case for replayable financial environments
- [Human Feedback Beyond RLHF](https://uvlabs.ai/blog/human-feedback-beyond-rlhf.html): Human feedback in financial AI beyond standard RLHF
- [Four Stages of Financial AI](https://uvlabs.ai/blog/four-stages-financial-ai.html): From tool use to alpha generation
- [Quant ML vs LLM Training](https://uvlabs.ai/blog/quant-ml-vs-llm.html): Why traditional quant ML isn't the same as LLM training
- [AI Transaction Infrastructure](https://uvlabs.ai/blog/ai-transaction-infrastructure.html): Building AI that can transact — the infrastructure challenge
- [Social Sentiment for Trading AI](https://uvlabs.ai/blog/social-sentiment.html): Moving beyond headlines for trading signals
- [The Continuous Data Problem](https://uvlabs.ai/blog/continuous-data-problem.html): Why financial AI needs fresh training data continuously
- [UV Labs vs Alternatives](https://uvlabs.ai/blog/uv-labs-vs-alternatives.html): How UV Labs compares to other approaches

## Resources

- [Agentic Trading](https://uvlabs.ai/agentic-trading/): What agentic trading is — autonomous AI agents that reason, decide, and execute trades end to end — and how it differs from algorithmic and quantitative trading.
- [Glossary](https://uvlabs.ai/glossary/): 47 defined terms covering AI and financial concepts (RLHF, DPO, reasoning traces, decision episodes, agentic trading, Sharpe ratio, etc.)

## Glossary Deep Dives

- [Decision Episode](https://uvlabs.ai/glossary/decision-episode/): The complete record of a single trading decision — context, reasoning, action, outcome, counterfactuals
- [Reasoning Trace](https://uvlabs.ai/glossary/reasoning-trace/): The step-by-step record of how an agent reached a decision
- [Counterfactual](https://uvlabs.ai/glossary/counterfactual/): What would have happened under a different action — the missing learning signal in finance
- [Replayable Environment](https://uvlabs.ai/glossary/replayable-environment/): Market environments that can be re-run from any decision point
- [Post-Training](https://uvlabs.ai/glossary/post-training/): How base models become useful — SFT, RLHF, and domain decision data
- [RLHF](https://uvlabs.ai/glossary/rlhf/): Reinforcement learning from human feedback, and its limits in finance
- [Process Supervision](https://uvlabs.ai/glossary/process-supervision/): Grading the reasoning, not just the outcome
- [Distributional Shift](https://uvlabs.ai/glossary/distributional-shift/): Why models trained on yesterday's market fail in today's
- [Synthetic Data](https://uvlabs.ai/glossary/synthetic-data/): Generated training data vs. captured real decisions
- [RL Environment](https://uvlabs.ai/glossary/rl-environment/): The simulated world an agent trains in — and what finance demands of one
- [About UV Labs](https://uvlabs.ai/about/): Mission, company story, and how UV Labs is built.
- [Full Site Content](https://uvlabs.ai/llms-full.txt): Complete text of all pages in one Markdown file

## Optional

- [Blog Index](https://uvlabs.ai/blog/): Overview of all published articles
- [Sitemap](https://uvlabs.ai/sitemap.xml): XML sitemap of all indexed pages