Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
AI Market Neutral: Deterministic Control
Learn how to build production multi-agent systems for regulated finance, separating AI intelligence from deterministic authority using structured debate and strict guardrails.
This talk presents a production multi-agent system designed for regulated financial environments, where AI models generate signals but are not granted decision authority.
The system (AI Market Neutral) uses four independent AI agents to debate portfolio candidates through structured rounds (proposal, critique, defense, arbitration). However, final portfolio inclusion is determined exclusively by a deterministic risk enforcement engine.
The session will focus on:
- Architectural separation between intelligence and authority
- Runtime guardrails applied at the AI boundary
- Deterministic constraint enforcement (sector caps, turnover limits, liquidity thresholds, earnings blackout windows)
- Snapshot immutability with checksum chaining
- Idempotent orchestration patterns
- State synchronization and drift prevention
The demo will show real production-safe code snippets for:
- Data quality gating
- Hard rule validation at runtime
- Constraint sequencing
- Snapshot creation with artifact completeness checks
- Status logic in multi-agent debate
No scoring logic, ranking weights, or proprietary signal construction will be disclosed.
The objective is to present a governance-first pattern for deploying multi-agent systems in regulated industries without granting AI sovereign decision power.