The AI Lab for FinTech
Self-Driving Alpha.
AI Agents That Trade.
We build autonomous AI that researches, trains, and deploys trading models around the clock — no human in the loop.
Experience
Trading
Operation
System
The Problem
Human traders lose to
their own psychology.
80% of retail investors underperform the market. Emotion, impatience, and attention-induced trading destroy returns. Algorithmic solutions exist — but they're black boxes that only provide signals, requiring humans to interpret and execute.
Our Solution
End-to-end agentic
intelligence.
Our Alpha engine doesn't just generate signals — it researches markets, trains reinforcement learning models, validates performance out-of-sample, deploys to live execution, and continuously self-improves. No human intervention required.
How It Works
Four stages. Fully automated.
Data Ingestion
Multi-source market data streams processed and normalized in real-time across asset classes.
RL Training
Reinforcement learning agents train continuously on GPU fleet, optimizing for risk-adjusted returns.
Validation
Walk-forward out-of-sample testing ensures no model goes live without proven performance.
Live Execution
Validated models deploy automatically. 24/7 monitoring with real-time risk management and position control.
Platform Architecture
Three layers. Fully agentic.
AI agents drive decisions. The pipeline trains and validates models. Proven models execute live — with performance feeding back into intelligence.
Closed-loop system: execution results drive continuous model improvement.
Multi-Tier Escalation
Intelligent cost control.
Lightweight agents handle routine monitoring. Complex decisions escalate to frontier models. Critical actions require human approval.
90% of events resolve at Tier 0–1. Escalation is automatic and cost-aware.
Our Edge
Built different.
Self-Evolving Models
Continuous reinforcement learning on a dedicated GPU fleet. Models don't just deploy — they retrain daily, adapting to market regime changes automatically.
Multi-Tier AI Escalation
Lightweight agents handle routine monitoring. Complex decisions escalate to frontier models. Critical actions require human approval. Cost-efficient by design.
Walk-Forward Validation
Every model is tested on unseen future data before deployment. No overfitting, no survivorship bias. If it doesn't prove itself out-of-sample, it doesn't trade.
Market Opportunity
Massive and growing.
Leadership
Built to win.
Leo Tang, PhD
Founder & CEO
19+ years of AI/ML R&D leadership at Google, Meta, LinkedIn, Microsoft, Amazon. Columbia PhD (CS). 10+ US patents, 20+ academia/research publications. Pioneer in agentic AI systems, multi-agent orchestration, reinforcement learning, and large-scale production AI Modeling. Spearheaded Google's first GenAI feature for App Search; built Meta's trillion-scale recommendation engine.
Shengbei Guo, MBA
President & CIO
29+ years in global trading and investment management across Wall Street and Asia. Morgan Stanley trader, Deutsche Bank MD (proprietary trading), CITIC Securities MD (alternatives & equities). PKU CS, Columbia MS, Wharton MBA. Founded GSB Podium Advisors.
Wharton
Interested in our
seed round?
We're raising seed funding to scale our autonomous trading platform. Let's talk.