Jordan Bee
Quant systems engineer. I build autonomous trading systems from scratch: signal generation, risk management, execution, monitoring. The systems, research code, and artifacts on this site are mine; simulations and archived results are labelled as such.
Skills
From signal generation to production execution.
- LLM orchestration and routing
- Multi-agent systems (Kaleo, 11 agents)
- Signal generation, regime detection
- Kelly criterion, fractional sizing
- Walk-forward backtesting, IC analysis
- Mean reversion + momentum strategies
Dedicated live host + AI hub, systemd fleets, Docker, WebSocket pipelines
Python, Elixir/OTP, REST APIs, SQLite, Postgres
SvelteKit, Three.js, GSAP, Tailwind, dashboards
TDD (5,500+ tests), CI/CD, security auditing
Why I Trade Algorithmically
I've been trading stocks and options manually for over 10 years. I've read the books, studied the charts, tried the strategies. And after a decade of it, I never found long-term, consistent success. Not because I didn't understand the markets, but because I couldn't get out of my own way.
The pattern was always the same. I'd have a good run, get confident, size up too fast, then give it all back in a few bad trades. I'd cut winners short because I was afraid of losing the gain. I'd hold losers too long because I couldn't accept being wrong. I'd overtrade after a win streak and freeze after a drawdown. Ten years of that.
Algorithmic trading removes me from the equation. The system doesn't feel fear after a drawdown. It doesn't get greedy after a streak. It executes the same strategy at 3 AM on a Sunday as it does at market open on a Monday. The edge isn't the strategy. It's the consistency of execution that no human can sustain.
A 49% win rate sounds terrible until you realize the TP:SL ratio is 1.8:1. The math works out to positive expected value on every trade. But only if you take every trade, with the correct sizing, without flinching. After 10 years of trying to do that manually, I know I can't. A machine can.
That's why I stopped trading manually and started building systems instead. The code is the discipline I never had. The validation harness is the evidence. And when the harness says a strategy is dead, it dies, no matter how much I liked it.
What I Bring to a Team
Research validation that kills its own ideas, execution and risk infrastructure that survives contact with real money, and agent systems that run unattended. Shipped live code, not just backtests.
End-to-end: signal research, position sizing, execution, and monitoring. Built and run live on Hyperliquid and Robinhood, at deliberately small size.
Kelly criterion sizing, circuit breakers, multi-server architecture, stop-loss layering. The guard rails that keep systems alive.
Deflated Sharpe, placebo nulls, walk-forward, power analysis. I have killed around twenty of my own strategies with this toolkit. I know what failure looks like before it happens.