I trade XAUUSD (gold / US dollar) full-time. Most of the quant and AI work on this site grows directly out of that practice — the systems I build are the systems I want to run or stress-test, not academic exercises.
Where the trading intersects the engineering
apes-signal is the clearest example. A TradingView Pine v5 momentum setup I’d been running, ported bar-by-bar to Python and hammered with an iterative ratchet optimiser against 125,000 hourly XAU/USD bars from 2004 to 2025. The final report says what the data said: the strategy had a ceiling below trading costs. Archived as a negative result rather than quietly shelved. That’s the kind of answer I build platforms to get — see Why I publish negative results.
FTR Strategy Backtester runs three systematic strategy families on a shared metrics spine — supply/demand FTR, order-block + trend filter, opening-range breakout. Confidence-adjusted expectancy, walk-forward validation, no-lookahead higher-timeframe bias, MD5-hashed data snapshots. The research platform is the product; the strategies themselves are commodities.
Eden Fintech Scanner (Python) sits at the other end of the spectrum — fundamentals on US-listed equities, weekly cadence rather than intraday XAU — but the discipline is the same. Hashed snapshots, contract-tested screens, LLM review behind typed information barriers, a written postmortem when things don’t work.
Stack
- Live: TradingView Pine v5 for prototyping and running setups
- Research: Python (pandas / NumPy on history-sized work; stdlib on production pipelines)
- Storage: Postgres for historical bars and run ledgers
- Optimisation: a home-brewed ratchet pattern — deterministic hypothesis catalogue, accept-or-revert, JSONL log, git commits for provenance
- Discipline: IS / OOS splits, walk-forward as a gate, confidence-adjusted metrics, data snapshot hashing
What I’ll take on
Systematic strategy research, signal pipelines, execution tooling, or a sanity-check on an existing backtest. The fuller picture is on Quant Engineering; the cover letter has the longer pitch; what I’m working on right now is where to look for current focus.