Override

Formula 1's 2026 regulation cycle turns energy management into a central strategic problem. The MGU-H is removed, MGU-K triples to 350 kW, and power delivery shifts to a 50/50 combustion-electric split. Existing racing AI tools were shaped around the 2014-2025 hybrid era and cannot provide an explainable layer for the new energy-management problem. Engineers need to understand why energy decisions make sense, what evidence supports them, and which regulation text they are grounded in.

Built a four-part proof frame: Detect (deterministic Python heuristics identify inefficient deploy/harvest/recharge zones), Explain (Granite 4 Instruct generates causal reasoning chains grounded in FIA regulations parsed by Docling), Validate (two-pass pipeline: deterministic checks + Granite Guardian BYOC scoring), Translate (Engineer Mode with reasoning chains and citations, Fan Mode for plain-language translation). The system supports counterfactual strategy review and optional 5-lap SoC forecasting via Granite Time Series TTM-R2.

End-to-end pipeline runs in ~8.2 seconds on watsonx.ai Essentials. 439 collected tests including watsonx integration tests. Ingests TORCS lab captures and FastF1-style replays. Generates regulation-grounded explanations with dynamic citations. Two-pass validation ensures energy-safety and regulation-consistency. Live TORCS cockpit integration for real-time strategy exploration.