Skip to main content
← All work
Case study 01

Investi

An autonomous trading desk built from three agents that disagree productively.

Year
2025 — present
Role
Founder & Product Manager
Stack
Python · OpenAI Agents SDK · PostgreSQL + pgvector · LangSmith · Alpaca API
Investi

The problem.

Most retail trading tooling is either a black-box robo-advisor or a screen full of charts. There's nothing in between for someone who wants logic-based, auditable strategies that they can read, debug, and trust.

The approach.

Three agents with different jobs and the right to overrule each other. The Researcher gathers market signals. The Analyst turns them into a position thesis. The Trader sizes and executes. Every step is logged in PostgreSQL with vector search over prior decisions, and every run is traced in LangSmith—so I can replay any trade and ask, honestly, why it happened.

The outcome.

A working autonomous investing system live at tryinvesti.co, executing logic-based strategies through Alpaca with persistent memory and observable reasoning. Still very much a work in progress—next up is a Risk Officer agent with veto power over the trader.

Next project OpenModel