Post-training tasks & data
Expert-built financial tasks and data, from core skills to complex workflows.
Hedge funds expertise built for AI training
We turn financial research workflows into
ready-to-use datasets, long-horizon training tasks, and reproducible evaluations.
- Long-horizon training tasks
- Reproducible evaluations
- Task-specific scoring
- Research and simulation tools
｜THE CHALLENGE｜
Real Financial Research Is Hard to Capture—and Evaluate.Model teams need realistic, long-horizon tasks,
expert research trajectories, and reliable ways to assess results.Creating them takes deep investment expertise and
tools to verify the work.
｜MISSION｜
CatPi Turns Years of Hedge Fund Experience into Model Capability.
Built by the team behind a leading quantitative hedge fund with a 14-year history,
we transform real investment research into training tasks, expert-reviewed evaluations,
and interactive learning environments.
｜Solutions｜
From expert-built tasks to scalable training environments.
Model evaluation & diagnostics
Reproducible evaluations to measure financial reasoning and pinpoint model weaknesses.
Financial RL environments
Proprietary backtesting, optimization and market simulation systems turned into RL environments
｜Multi-Asset · Multi-Strategy｜
From investment strategies to AI training.
We turn real-world strategy research into financial AI training tasks—from signal discovery
and portfolio construction to backtesting and risk analysis.
｜CORE TEAM｜
CatPi was incubated by a leading quantitative hedge fund with
a 14-year history. We develop training data, evaluation platform, and reinforcement learning environments for financial AI.
Our core team includes graduates of MIT, Tsinghua, USC, and Columbia, with expertise in quantitative and fundamental research, AI, systems engineering, and data science.
Bring Investment and Trading Expertise
to Your Model Training.
Explore sample research tasks, expert-reviewed reference outputs, and evaluation criteria grounded in real investment workflows.