AI/ML Prediction Market Developer

Hi, I’m looking for an experienced AI/ML developer to build a sophisticated real-time prediction market AI for **ᴋᴀʟꜱʜɪ ᴀɴᴅ ʀᴏʙɪɴʜᴏᴏᴅ 15-ᴍɪɴᴜᴛᴇ ᴘʀᴇᴅɪᴄᴛɪᴏɴ ᴍᴀʀᴋᴇᴛꜱ**. This is intended to be a serious, fully tested system—not a basic indicator, chatbot, or simple prediction script. The goal is to create an AI that continuously monitors live markets, analyzes historical data, scans relevant live news, applies multiple trading/prediction strategies, and only gives a trade signal when the evidence meets strict confirmation requirements. I understand that no AI or trading system can guarantee a perfect or near-perfect win rate. I am not looking for unrealistic promises. I want the system engineered to maximize accuracy and have its actual performance proven through rigorous testing. ### 1. LIVE MARKET DATA The bot should continuously monitor available live data from Kalshi and, where technically and contractually supported, Robinhood. It should analyze information such as: * Current price * Bid/ask * Market-implied probability * Volume * Liquidity * Order book/depth where available * Price movement * Momentum * Market structure * Spread * Trading activity * Time remaining * Other relevant real-time variables It should continuously monitor the market rather than simply taking one snapshot every 15 minutes. Timing is extremely important because these markets can change rapidly near expiration. ### 2. HISTORICAL DATA & BACKTESTING The AI should analyze as much reliable historical data as reasonably available. I want a historical database for: * Pattern recognition * Statistical analysis * Strategy testing * Machine-learning training * Historical probability analysis * Comparing current conditions with similar historical situations * Backtesting The system must properly separate training and testing data and prevent **data leakage and overfitting**. Testing should include: * Training/validation data * Out-of-sample testing * Walk-forward testing * Historical backtesting * Forward testing * Live paper trading I want performance measured on data the model did not use during training. ### 3. LIVE NEWS & EXTERNAL DATA The AI should continuously scan relevant live information that could affect a specific market, including: * Breaking news * Economic announcements * Government releases * Financial news * Major events * Weather where relevant * Company announcements * Economic indicators * Scheduled events * Other market-moving information The AI should not simply display headlines. It needs to determine whether the information is actually relevant, whether it could change the probability of the outcome, how significant the impact may be, and whether it supports or conflicts with the current market signals. If news creates uncertainty, the system should be capable of saying **NO TRADE**. The news and external information considered for each prediction should be logged so we can later determine how it affected the result. ### 4. MULTI-STRATEGY CONFIRMATION I do not want trades based on one indicator or one AI prediction. The system should use multiple independent factors and strategies, potentially including: * Price action * Momentum * Volume * Market structure * Liquidity * Order-book behavior * Short-term trends * Mean reversion * Breakouts * Statistical probability * Historical patterns * Market-implied probability * Technical indicators where appropriate * News sentiment * News impact * Event-driven analysis * Time-to-expiration behavior * Other statistically validated strategies I want these factors combined into a **structured confirmation system**. For example, if historical patterns, momentum, market structure, volume, current probability, and relevant news all support the same outcome, confidence may increase. If important signals conflict, the system should reduce confidence or return **NO TRADE**. The confirmation system must be based on measurable historical performance—not an arbitrary score designed to make the AI look sophisticated. ### 5. CONFIRMED TRADE CALLS When a trade meets the confirmation requirements, the bot should clearly provide: **TRADE CONFIRMED** * Market/contract * Direction/outcome * Current price * Recommended entry price/range * Signal time * Time remaining * Confidence/probability estimate * Historical probability * Supporting strategies/signals * Relevant news * Reason for confirmation * Expected value/risk information where appropriate * Conditions that would invalidate the signal If the evidence is insufficient: **NO TRADE** I want the AI to be selective. I would rather receive a small number of highly qualified signals than many weak predictions. ### 6. PERFORMANCE TRACKING Every prediction must be automatically recorded. The system should track: * Date/time * Market * Prediction * Entry price * Final result * Confidence * Strategies used * News considered * Market conditions * Whether the trade was confirmed * Whether the prediction was correct * Simulated profit/loss * Reason for the decision It should calculate: * Total signals * Win rate * Loss rate * Accuracy * Performance by market * Performance by strategy * Performance by confidence level * Performance during news events * False-positive rate * Missed opportunities * Simulated P/L * Expected value * Drawdown * Performance over time * Performance under different market conditions A dashboard/reporting system showing these statistics would be preferred. ### 7. PAPER TRADING Before using real money, I want a **paper-trading/live simulation mode**. The bot should use live market and news data and record the trades it would have taken without risking real money. This is important because historical backtesting alone is not enough. I want to evaluate how the system performs with real-time data and real-time decisions. ### 8. MACHINE LEARNING If machine learning is used, please explain: * What models you would use * What data/features would be used * How the model would be trained * How validation would work * How