Python developer — crypto paper-trading harness

Single-process Python 3.11 harness for an automated BTC system, delivered in stages as one package. A separate strategy module (provided) is called through a defined interface and tested against an included mock. Interface spec and tests are in the brief. Stage 1 — Paper-trading harness - APScheduler: pull BTC/AUD 4h candles from Kraken via ccxt, hourly - Indicators: EMA21, EMA55, RSI14, 20-bar swing high/low - Paper simulator: spot fills + 2:1 CFD positions with margin tracking, liquidation price, daily funding accrual - SQLite ledger: trades, positions, bank, accruals — append-only - Alerts + remote control: push notifications (fills, margin events) and remote /status and /kill via a messaging bot API (platform named in brief) - YAML config - pytest coverage on simulator and ledger - Runs unattended on a basic Ubuntu VPS, one process, 72h zero crashes, against the mock strategy Stage 2 — Integration - Wire the provided strategy module into the harness through the interface - Resolve harness-side fixes surfaced during integration Stage 3 — Live execution + accounting (after Stage 1+2 accepted) - Live adapters: Kraken live keys via ccxt; IG REST/streaming for CFD legs (confirm AU-retail BTC CFD availability; Pepperstone MT5 fallback) - Live accounting: real fills + funding to ledger; nightly broker-statement reconciliation; quarterly/EOFY CSV export - Go-live wiring at floor size CONSTRAINTS - One process on a basic Ubuntu VPS. No Docker, no microservices, no dashboard. - Work-for-hire, full ownership on payment. Private repo. NDA required. - Acceptance is test-based: tests provided in the brief, a stage passes when its tests pass. Full brief shared on engagement after NDA. BUDGET: US$1,000 fixed, staged for tracking, released on accepted delivery. TO APPLY 1. How would you structure the CFD margin simulation? 2. Realistic total effort estimate for Stage 1. 3. How you use AI coding tools day to day.