Processing gigabytes of live F1 telemetry through Next.js Server Components. Gemini LLM inference combined with a MongoDB memory loop that compounds accuracy across every race weekend.
Tyre degradation slopes mapped per compound. Crossover lap calculations factor in pit-lane time loss and tyre warm-up phase.
"Norris must pit by Lap 22. Beyond this, the hard tyre warm-up phase will allow Verstappen to execute the undercut through Turns 4–7."
"The rear is very loose under braking for Turn 6. I need more front wing, maybe half a turn."
"Tyres are starting to come in. The front feels connected now, let’s push this lap."
"Getting some oscillation out of Turn 12. Is it the active aero mode switch timing?"
Server Actions aggregate, down-sample to 1Hz, and cache processed summaries in MongoDB before the client ever makes a request. Sub-100ms load times with zero risk of OpenF1 rate limiting.
{
"_id": ObjectId("67a9f3..."),
"race_id": "2026-AUS-R01",
"session": "Qualifying",
"predicted": [
"VER", "NOR", "PIA",
"LEC", "HAM"
],
"actual": [
"VER", "PIA", "NOR",
"HAM", "LEC"
],
"accuracy": 0.72,
"retrospective":
"Under-weighted Piastri
X-Mode efficiency on
Albert Park main straight.
Adjust low-drag circuit
weighting by +15%."
}