Car race Sync

Inspiration

Every driver has the same problem: how do I improve faster? Traditional coaching is expensive and reactive. We realized telemetry data holds all the answers—throttle patterns, brake fade, lap inconsistencies—but drivers can't analyze it in real-time. What if an AI coach could instantly identify weaknesses, suggest micro-adjustments, and track skill progression across every race? RaceCraft transforms raw telemetry into personalized coaching, making professional-level analysis accessible to all drivers.

What It Does

RaceCraft is an AI-powered motorsports coaching platform that guides drivers through four critical phases:

• Pre-Race: AI Coach analyzes driver history, track characteristics, and competitor data to generate personalized strategy documents with focused throttle/brake targets
• During Race: Real-time telemetry monitoring detects brake fade patterns and throttle spikes, sending context-aware alerts with micro-adjustment suggestions
• Post-Race: Lap replay analysis, head-to-head competitor comparisons, and AI-generated improvement drills with measurable goals
• Lifecycle Tracking: Persistent driver profiles track skills (throttle control, braking precision, racecraft) with progression curves and achievement milestones

Every race feeds into the driver's profile—creating continuous, data-driven improvement.

How We Built It

Backend: Python-based AI Coach using Azure OpenAI GPT-4 for intelligent analysis and coaching generation. Context injection injects driver telemetry, track data, and competitor metrics into coaching prompts.

Frontend: Streamlit dashboard displaying driver metrics, AI chat interface, lap comparisons, and progression visualizations.

Data Pipeline: Consolidated telemetry processing across 12 race events (Barber, COTA, Road America, Sebring, Sonoma, VIR), computing lap summaries, performance metrics, and driver statistics.

Architecture: Backend/frontend separation with analytics engine computing real-time detection pipelines and comparative analysis matrices.

Challenges We Ran Into

• Real-Time Detection Accuracy: Distinguishing genuine brake fade from normal braking variation required statistical analysis of brake pressure curves, temperatures, and lap patterns
• Context Injection: Balancing telemetry data richness with API token limits while maintaining coaching quality
• Data Consolidation: Normalizing telemetry across different car setups and track configurations
• User Trust: Ensuring AI suggestions were specific, actionable, and based on objective data—not generic advice

Accomplishments We're Proud Of

• Built an intelligent coaching system that understands driver context and tailors advice to individual needs
• Developed real-time anomaly detection that works during live races
• Created a lifecycle progression system that motivates drivers with measurable growth metrics
• Achieved clean backend/frontend architecture enabling easy feature additions
• Processed 12 complete race datasets with consistent, normalized metrics

What We Learned

• AI Context Matters: Generic coaching is useless; injecting driver data transforms responses from generic to personally relevant
• Telemetry is Gold: Drivers' instincts often conflict with data; objective analysis builds trust
• Consistency Drives Growth: Tracking skills across multiple races reveals patterns invisible in single-race analysis
• Simplicity Scales: Focused, clean architecture beats feature-heavy complexity for real-world deployment

What's Next for RaceCraft

• Mobile App: Push notifications for real-time alerts during races
• Multi-Driver Teams: Head-to-head driver development within racing teams
• Predictive Analytics: AI predicting optimal pit stops, fuel strategies, and tire management
• Voice Coaching: Real-time audio guidance through car communication systems
• Setup Optimization: AI-driven suspension/brake balance recommendations based on driver telemetry patterns

Built With

• car
• llm
• race
• streamlit