Specialist AI trading research and machine-learning development project

# Senior Quantitative Trading AI Engineer — Market Structure Recognition, Historical Learning & Live Trading Research Agent **Experience Level:** Expert **Project Type:** Custom AI / Machine Learning / Quantitative Trading Software **Engagement:** Initial paid milestone, with potential for long-term development **Compensation:** Competitive, based on experience and technical proposal ## Project Overview I am seeking an experienced quantitative trading software engineer and AI developer to build a specialised AI-powered trading research and learning system. This is **not a standard trading bot, indicator-based signal generator or basic AI chatbot**. The objective is to develop an intelligent research platform capable of analysing historical forex and gold price data, recognising complex market structures, identifying recurring price-action sequences, comparing successful and unsuccessful historical setups, and progressively improving its recognition capabilities through validated learning. The system will eventually monitor live market conditions and identify potential high-quality trading opportunities for human review. The initial development will focus exclusively on **XAU/USD (Gold)**, with the architecture designed to accommodate GBP/USD and GBP/JPY later. I already have an established discretionary trading methodology. The developer's role is to translate this methodology into a reliable, testable technical system—not invent a new trading strategy. ## Core Technical Requirements ### 1. Historical Market Data Develop a reliable data pipeline for historical XAU/USD price data across: - Daily (D1) - 4-hour (H4) - 1-hour (H1) - 15-minute (M15) - 5-minute (M5) Experience with MetaTrader 5, Python and reliable historical market-data providers is highly desirable. The system must correctly handle candle timestamps, multiple timeframes, missing data and historical data integrity. ### 2. Advanced Market Structure Recognition Develop algorithms capable of identifying and analysing: - Bullish and bearish trends, channels and ranges - Compression and consolidation - Trend exhaustion and weakening momentum - Break of Structure (BOS) - Change of Character (CHOCH) - Liquidity sweeps and equal highs/lows - Supply and demand zones - Mitigation and structural retests - Double tops, double bottoms, flags, wedges and other relevant price formations - Fibonacci confluences using a predefined set of approved levels - Multi-timeframe structural alignment The system must analyse the **sequence and development of price behaviour**, rather than relying exclusively on isolated candlestick patterns. ### 3. Historical Pattern Discovery and Similarity Search The system should: - Analyse approximately 12–24 months of historical data initially. - Identify recurring market structures and price-action sequences. - Retrieve comparable historical setups. - Distinguish successful, unsuccessful and ambiguous examples. - Examine what happened after each identified structure. - Explain which structural characteristics appear relevant to the outcomes. Historical similarity must be based on information available at the relevant time, without using future outcomes to select examples. ### 4. AI Learning and Model Improvement Develop a controlled learning process that can: - Learn from labelled historical chart examples. - Incorporate human feedback and corrections. - Improve market-structure recognition. - Compare updated models against previous versions. - Identify potentially useful recurring structural characteristics. - Evaluate improvements using previously unseen historical data. The system must not automatically modify the approved trading strategy or live trading rules. We are particularly interested in genuine machine-learning capability rather than simply adding an LLM interface to conventional trading software. ### 5. Historical Backtesting and Validation Implement reliable testing procedures including: - Chronological historical simulation - Prevention of look-ahead bias and repainting - Realistic transaction costs and execution assumptions - Out-of-sample evaluation - Walk-forward validation - Recognition accuracy, precision, recall and false-positive measurements - Trading expectancy, drawdown and risk/reward analysis where applicable - Version-to-version performance comparisons All findings should be reproducible and supported by clear evidence. ### 6. Live Market Monitoring — Later Development Phase Following successful historical validation, the system should be expandable to: - Monitor live XAU/USD market data. - Identify developing structural setups. - Distinguish developing, confirmed and invalidated conditions. - Highlight potential precision-entry opportunities. - Calculate potential structural stop-loss, invalidation, risk/reward and position sizing. - Generate alerts with explanations and relevant historical comparisons. **Automatic live trade execution is not required for the initial project.** ### 7. Research Interface and Explanations The system should ultimately provide: - Annotated historical charts - Recognised structural sequences - Similar historical examples - Outcome comparisons - Clear explanations of the evidence supporting each detection - Research reports and performance metrics A simple functional interface is acceptable initially. The priority is accurate research and recognition, not an elaborate dashboard. ## Required Developer Experience Applicants should demonstrate strong experience in several of the following areas: - Python software engineering - Quantitative trading research - Forex and/or gold market data - Market-structure and price-action algorithms - Historical backtesting and validation - Machine learning and pattern recognition - Time-series analysis - MetaTrader 5 integration - Historical similarity search - AI agents and LLM integration - Production-quality data pipelines and software architecture Experience with Claude Code, LangGraph or similar agent-development tools is welcome, but is not a substitute for quantitative trading expertise. ## Development Approach — Milestone Based I intend to develop the platform progressively, with each milestone requiring a demonstrable and testable deliverable before further development is authorised. **Milestone 1 — Technical Discovery and Architecture** Review the existing trading methodology and sample charts. Deliver a technical design, proposed architecture, data-source recommendations, initial recognition definitions and a realistic implementation plan. **Milestone 2 — Historical Data Foundation** Develop and validate the XAU/USD historical data pipeline. **Milestone 3 — Market Structure Recognition Prototype** Implement and demonstrate recognition of a small number of predefined price-action structures. **Milestone 4 — Historical Similarity Research** Retrieve and compare structurally similar historical examples. **Milestone 5 — Independent Testing and Handover** Demonstrate recognition performance on unseen examples, document limitations and provide complete source code and technical documentation. Later milestones may include adaptive learning, advanced backtesting, live scanning and automated alerts. The initial engagement will begin with **Milestone 1 only**. Continued development will depend on demonstrated technical capability and successful milestone delivery. ## Application Questions Please answer the following questions in your proposal: 1. Have you previously developed quantitative trading systems involving forex, gold or market-structure analysis? Please provide specific examples. 2. Have you programmed BOS, CHOCH, liquidity sweeps, supply/demand, channels or other advanced price-action structures? Explain your approach. 3. What experience do you have with historical XAU/USD data, MetaTrader 5 and multi-timeframe analysis? 4. How would you design a system to find historically similar price-action sequences rather than simply matching candlestick shapes? 5. How would you implement feedback-driven learning while preventing overfitting, look-ahead bias and deterioration in model performance? 6. What technologies and architecture would you recommend for this project, and why? 7. Can you demonstrate an existing trading research, backtesting or market-structure recognition project you have developed? 8. What deliverables, estimated hours and fixed-price quotation would you propose for the initial technical-discovery milestone? ## Important I am looking for an experienced engineer who can demonstrate genuine expertise in quantitative market research and AI development. Please do not submit generic AI chatbot proposals, basic indicator-based trading bots or claims of guaranteed trading profitability. Strong communication, reproducible technical work, milestone accountability and complete source-code handover are essential. **The objective is to build a serious, evidence-based AI trading research system that becomes more effective through measurable, validated learning.**