Daily Profit Model
This Ospina case study documents how Carlos Rico-Ospina approached a specific risk, infrastructure, revenue, or research problem and what was built to address it.
LightGBM forecasting model for day-ahead P&L prediction—research project exploring predictive risk signals.
The Problem
Trading operations produce huge volumes of event-level data, but decision-making often happens at the wrong granularity. The KPI that matters operationally is daily P&L, not individual trades.
The Insight
Daily profit is a 'system-level output' of many micro-decisions. If you can forecast D+1 profit at the account level, you can drive risk controls, monitoring, and operational interventions earlier.
What I Built
- Implemented LightGBM forecasting model for day-ahead profit prediction
- Designed pipeline compatible with large datasets (tens of millions of trades)
- Built repeatable feature generation and training loop with leakage-safe design
Outcomes
- Delivers practical 'tomorrow P&L' signal upstream of risk controls
- Supports anomaly detection for accounts diverging from expected performance
- Research project demonstrating end-to-end forecasting system design
Related Projects
View All
ToolsTrade Alerts
2025Code-defined FX monitoring with push notifications—because alerts should be reproducible, not UI-configured.
ToolsCarbitrage
2025Weekend prototype for car flipping arbitrage—aggregates dealer offers and computes deal viability in real-time.
Have a Similar Challenge?
Let's discuss how I can help you achieve similar results.
Start a Conversation