Founder-Engineer
For two years I built and ran ShareShark solo, a dual-currency sweepstakes platform where users predict stock price movements. I built the whole thing: a calibrated ML pricing engine using live market data, a suite of AI risk agents guarding the platform, a KYC/AML/geofencing compliance stack, and the money ledger. ShareShark ran in production through a year of free-to-play and a real-money soft launch with ACH payouts.
I think in terms of expected value, probability, and risk management. I pick up new domains fast, and work comfortably in niche or heavily regulated industries.
B.B.A. Finance, magna cum laude, University of Georgia.
What I build
Money-correct systems where every rail fails closed: ACH payouts, KYC/AML, geofencing & fraud, encryption, and W-9/1099 tax reporting.
Production-grade AI: multi-agent systems, cost-tiered LLM routing, model output treated as untrusted, and full decision audit trails.
Calibration-first ML pricing (beats Black-Scholes), Student-t copula Monte Carlo, fractional Kelly sizing, out-of-sample edge validation, and model-free risk bounds.
Case studies
A LightGBM pricing engine with calibration-first evaluation and heavy overfitting controls: Bayesian tuning, out-of-sample backtesting, and guards that auto-reject bad models. The result is calibrated ~19× tighter than Black-Scholes, the Nobel Prize-winning options formula.
Read the case study → Data Engineering · MLEveryone obsesses over the model; the data feeding it is what actually makes or breaks it. I pulled eight years of messy data from six data sources and lined it up so the model could never peek at the future: the leak that makes a backtest look brilliant and the live model near-useless.
Read the case study → Quant · MLSeveral times I tried to improve the pricing model with more complexity: more models, more data, more specialization. Honest backtests overruled me, so I shipped the single, simpler model and deleted the rest.
Read the case study → Quant · MLWhen users combined several predictions into one all-or-nothing entry, multiplying the odds together quietly overpaid them: three tech stocks all 'higher' is essentially one entry, not three. I priced the true joint odds with a copula simulation that models how stocks move together, and it stayed margin-positive across a six-year backtest.
Read the case study →When news breaks while the market is closed, a stale price is free money for whoever notices first. Five agents watch the news and trading flow 24/7 to close that window. Two more run ops and growth, and every agent's output is validated before it can move a price.
Read the case study → Regulated-FintechThe full money-and-compliance backend for a real-money platform: bank payouts, identity checks, fraud, and taxes. Money only moves once every check passes, and a payout never goes out twice.
Read the case study →About
I'm a founder-engineer who builds production systems, from data pipelines and ML models to live serving and safety rails. Most of that came from building ShareShark solo, a dual-currency sweepstakes platform where users predict stock price movements. I built a concurrency-safe money ledger, an ACH payout and KYC/AML/tax compliance stack, a calibrated LightGBM pricing engine, a suite of cooperating Claude-powered risk agents, and a cross-platform Vue/Capacitor client.
It ran in production through a year of free-to-play and a limited real-money soft launch, and it stays live as free-to-play today. I bootstrapped ShareShark with my own independent quantitative research across skill-based prediction contests and prediction markets, building specialized calibrated models and a risk toolkit with disciplined position-sizing. Finance degree, pricing & valuation emphasis (UGA, magna cum laude).
Contact
Open to roles in backend, payments, and fintech engineering; applied-AI / forward-deployed; ML / data engineering; quant / pricing & risk; and finance / valuation. I'm especially drawn to early-stage and founding-engineer roles where I'd own a lot of the work.
cooper@coopernorman.dev · linkedin.com/in/wcoopernorman · github.com/coopernorman
Download résumé (PDF) · See ShareShark live (shareshark.app), test login on request