Fix an existing Kalshi trading bot (TypeScript/Node) — paid on proven paper-mode profitability

We have an automated trading bot built in TypeScript/Node that trades on Kalshi, a regulated prediction market. It trades Kalshi's 15-minute Bitcoin up/down contracts (binary YES/NO markets that resolve every 15 minutes). The bot watches the market, buys into the favored side within a set price band late in each window, and manages the position with a stop-loss. The bot is built and running in paper-trading mode. The problem: it runs, but it is not producing a profit. We're looking for someone who knows algorithmic trading and trading-bot mechanics (entry/exit logic, position management, execution timing, risk controls) to find why it loses money and fix the trading mechanics so it's profitable in paper mode. You don't need Kalshi experience, you need trading-bot experience. You'll get access to the GitHub repository, the deployment, and the logs. Timeline: 1-2 weeks. Fixed price, single milestone. Payment is released once the bot is shown profitable in paper mode. Definition of done: - Fixes are committed to the repository with a short summary of what changed. - The bot runs for 48 continuous hours in paper mode. - Net P&L over those 48 hours is positive after fees, shown in the bot's trade logs. If this goes well, there's potential follow-on work as the bot moves to live trading.