Meet the author

Hi, I'm Victor Vega Sobral (a.k.a. VforVitorio). I graduated in Intelligent Systems Engineering from UIE Campus Coruña (9.0/10 average, four Distinctions / Matricula de Honor) and I'm now a Junior AI Engineer at NTT DATA Spain, working across computer vision, LLM behavior, agentic AI, and tooling such as Model Context Protocol (MCP) servers. Alongside the role I'm pursuing a Master's in Artificial Intelligence Research (AEPIA / UIMP). F1 StratLab was my Final-Degree Project, defended in June 2026 with a 10/10 grade and Distinction (Matricula de Honor) and unanimously recommended by the tribunal for publication as a research article: an open-source multi-agent AI for real-time Formula 1 race strategy, built end-to-end from telemetry ingestion and ML modelling to a LangGraph orchestrator and three operator surfaces. I built it to see how far a single thesis could push a digital twin of an F1 race.

Where to find me

• GitHub, @VforVitorio
• Project landing, f1stratlab.com
• DeepWiki, F1 StratLab on DeepWiki
• LinkedIn, victorvegasobral
• Hugging Face dataset, f1-strategy-dataset
• Portfolio, victorvegasobral.com

About the project

F1 StratLab is a multi-agent system that turns a live (or replayed) Formula 1 race into actionable strategy. Seven ML models cover the quantitative core, lap-time delta, tire degradation with MC Dropout, overtake probability, safety-car prior, pit duration and undercut success, and feed six LangGraph sub-agents (pace, tire, race situation, pit strategy, radio, RAG) which a single orchestrator (N31) fuses into a Pydantic-typed decision per lap: action, pace target, risk level, and a plan for the next pit window.

The same engine drives three operator surfaces: a React web app for analysts (backed by a FastAPI/MCP API for programmatic access and chat tool-calling), a CLI for headless replays, and a three-window arcade (race replay, strategy dashboard, live telemetry) built in PySide6 + pyglet for the demo experience. The whole stack is open under Apache-2.0 and shipped as wheels and GitHub releases through release-please automation.

This documentation site is the engineering companion to the thesis memoria, every notebook, model, agent and surface is wired into the graph view so you can navigate by topic, by tag, or by cross-reference.

Acknowledgements

• Academic: UIE Campus Coruña, Intelligent Systems Engineering faculty and thesis advisors.
• Open data community, FastF1 and OpenF1 provide the telemetry, lap and timing data this entire project depends on.
• Open-source libraries: LangGraph, LightGBM, XGBoost, PyTorch, Pydantic, FastAPI, React, Vite, ECharts, PySide6, pyglet, Qdrant, and the Hugging Face ecosystem.
• Reference work: TUMFTM race-simulation for the pit-delta framing, plus the wider F1 analytics community whose public notebooks shaped the early modelling decisions.

No copyright infringement intended. Formula 1, F1, and related marks are trademarks of Formula One Licensing B.V. and are used here for reference only. This project is not affiliated with, endorsed by, or in any way officially connected to Formula 1, the FIA, or any F1 team.