Systems across trading, execution, research, and financial modelling.

Context

Trading strategies often evolve into monolithic scripts where signal logic, position sizing, evaluation rules, order handling, and diagnostics are tightly coupled. That makes optimisation misleading, live failures harder to diagnose, and reuse expensive.

Approach

Design a reusable architecture in NinjaTrader 8 with independent modules, configurable profiles, explicit controls, and Strategy Analyzer compatibility.

Key contributions

• Modular strategy interfaces
• Separation of strategy, execution, risk, evaluation, connectivity, and reporting
• Execution and order-management abstraction
• Dynamic position sizing
• Volatility-based risk management
• Configurable prop-firm profiles
• Daily-loss controls
• Intraday and end-of-day trailing drawdown logic
• Profit-target and consistency rules
• End-of-day flat logic
• Automatic restrictions when evaluation constraints are violated
• Connection and market-state safeguards
• News and event filtering
• Market-regime filters for session, volatility, and trend
• Structured logs, diagnostics, and debugging support
• Parameter exposure for large-scale optimisation workflows

Outcome

Created a reusable system foundation for researching, validating, and deploying multiple strategies under consistent operational controls.

Key contributions

• Systematic strategy design
• Strategy logic and risk-rule implementation
• Multi-year historical research
• Large-scale parameter optimisation
• Robustness and market-regime review
• Execution frameworks
• Dedicated infrastructure for continuous operation
• Reconnection handling
• Data-interruption handling
• Infrastructure-failure safeguards
• Monitoring and diagnostics
• Statistical performance analysis
• Iterative model refinement

Outcome

Established a repeatable path from research to monitored live execution, with explicit attention to operational failure modes rather than backtest results alone.

Context

Simple order duplication is insufficient when accounts disconnect, fills differ, positions diverge, or commands arrive in an unexpected state.

Key contributions

• Real-time trade replication
• Pre-trade account and position checks
• Position verification during entry, management, and closure
• Discrepancy detection
• Slippage and tolerance thresholds
• Alerts and logging
• Account-state monitoring
• Desynchronisation prevention controls
• Recovery-oriented diagnostics
• Continuous refinement of execution logic

Outcome

Designed the system around state consistency, exception visibility, and recovery controls rather than simple order duplication.

Key contributions

• Data ingestion and preprocessing
• Feature engineering and feature selection
• Model training, validation, and evaluation
• Classification for directional signals
• Short-horizon predictive modelling
• Backtesting with transaction costs and slippage
• Execution logic and trade simulation
• Dynamic position sizing based on confidence, volatility, and risk constraints
• Trade journaling
• System logging

Outcome

Structured machine-learning experimentation inside a modular, testable workflow rather than treating a predictive model as a complete trading system.

Key contributions

• End-to-end technical development support
• Core rating logic in Python
• Financial modelling and data-driven analysis
• Snowflake backend infrastructure
• Posit Cloud environment
• Analyst-facing R Shiny interface
• Python/R integration through reticulate
• Automated data acquisition and preprocessing
• Backend-to-frontend integration
• Architecture, user-guide, and technical-documentation work

Outcome

Combined quantitative modelling, scalable data infrastructure, and analyst-facing workflows in one integrated financial system.

What remained confidential

• No bank name
• No internal model equations
• No client data
• No internal screenshots
• No claim of sole ownership of work completed within a wider team

Key contributions

• Training materials on IFRS 9, credit modelling, and statistical computing in R
• Stakeholder training support
• Modelling and segmentation discussions
• Definition of target modelling approaches
• Data-quality tests aligned with recognised European banking dimensions
• Data preparation
• Risk-driver selection
• Logistic regression fitting
• Calibration
• Backtesting

Outcome

Applied model-development discipline, data-quality controls, and stakeholder communication in a supervisory context.

Key contributions

• Validation-standard drafting
• Review of governance and model-risk-management processes
• Definition of validation tests and thresholds
• Independent challenge of model-development decisions
• Rating-system scope and risk-differentiation analysis
• CRR III interpretation and peer benchmarking

Outcome

Strengthened the validation framework around governance, regulatory expectations, risk differentiation, and independent challenge.

Key contributions

• Market-data ingestion
• Technical indicators
• Classification and regression models
• Rule-based and model-driven strategy components
• Portfolio creation and supervision
• Expected-return and profitability analysis
• Broker costs, commissions, and tax considerations
• Volatility measures and risk-adjusted evaluation
• Expandable model architecture
• Graphical user interface
• Supervised and automation-oriented workflows
• Broker API integration design

Academic proof

• Combined project across both degrees
• Technical and economic aspects assessed separately
• Experimentally oriented development with practical market application
• Academic supervision and independently evaluated technical and economic components