X-RaceIQ

X-RaceIQ puts a world-class AI race engineer in your helmet—real-time insights, instant tactical coaching, and smarter, safer, faster laps powered by autonomous analytics. The future of racing.

Comment

Inspiration

Racing demands split-second decisions and perfect awareness of machine and environment. We built X-RaceIQ: our digital platform and voice-activated AI race engineer that lives in your helmet. The AI Crew Chief Radio hardware device delivers real-time telemetry analysis and tactical coaching in under 2 seconds, while the HeartBeat Dashboard coordinates 8 autonomous agents to provide world-class engineering intelligence (👉 CLICK HERE FOR APP DEMO - link):

Image

Image

👇 Click image to watch driver video Watch Video

What it does

The X-RaceIQ Platform provides professional-grade, real-time telemetry analysis and strategic coaching through eight autonomous agents. It monitors:

1. Car Telemetry Agent
Responsible for loading race data from CSV, tracking current position (P2), and monitoring gap to leader (1.26s)

2. Pace Regression Agent
Responsible for analyzing the last 5 laps to identify performance trends and consistency patterns

3. Weather Agent
Responsible for monitoring track temperature (18.2°C) and calculating grip score (85%) based on environmental conditions

4. Engine/Tire Agent
Responsible for predicting tire wear (67%) and estimating remaining tire life (5 more laps)

5. Driver State Agent
Responsible for monitoring driver biometrics including heart rate (158 bpm) and fatigue levels (38%)

6. Lead Chase Agent
Responsible for calculating required pace to catch the leader (-0.16s/lap) and determining if overtaking is possible

7. LLM Insight Agent
Responsible for calling OpenAI/Claude with full race context and generating natural language coaching messages

8. Heartbeat Dashboard
Responsible for combining all agent outputs and making strategic race decisions via Heartbeat Dashboard

How we built it

We combined embedded hardware, on-helmet audio feedback, sensor fusion, and a multi-agent AI architecture. Each agent specializes in a domain: performance, strategy, health, or environment, and syncs data with our HeartBeat dashboard for live team analytics.

Complete System Integration Guide

AI Chief Radio Device + Agent2Agent API Platform

🎯 Overview: How Hardware and Software Work Together

There are TWO complete systems that work together:

  1. Agent2Agent API (Software Backend) - The intelligence
  2. AI Chief Radio (Hardware Device) - The communication interface

Here's how they integrate:

System Architecture: The Data Lap

Phase 1: In The Race Car (Edge)

  • Hardware: Driver Helmet + Bone Conduction Mic + Jetson Nano
  • Action: Driver speaks: "Hey Chief, what's my gap?"
  • Edge Processing: Local Speech-to-Text conversion (Whisper) happens on the Jetson Nano (~300ms).
  • Transmission: Text payload sent via 5G/WiFi to the Cloud.
⬇️ 5G Transmission ⬇️

Phase 2: Agent2Agent Swarm

The server receives the query and works with up to 8 Agents in Parallel to analyze the race state instantly and provide feedback to the Heartbeat dashboard and driver.

Agent Name Role & Data Source Insight Generated
1. Telemetry Loads CSV Race Data Current Pos: P2 / Gap to Leader: 1.26s
2. Pace Regression Analysis (Last 5 laps) Trend: Consistent pace
3. Weather Environmental Sensors Track Temp: 18.2°C / Grip: 85%
4. Engineering Vehicle Health Monitor Tires: 67% (Good for 5 more laps)
5. Driver State Biometrics Heart Rate: 158 bpm / Fatigue: Low
6. Lead Chase Strategy Calculation Required Pace to pass: -0.16s/lap
7. LLM Insight Contextual Reasoning Calls LLM to synthesize coaching message
8. Heartbeat The Decision Maker Combines all outputs into final strategy
⬇️ Total Latency: < 2 Seconds ⬇️

Phase 3: The Advantage (Feedback)

  • Payload: Text response returns to the Jetson Nano.
  • Output: Text-to-Speech synthesis generates audio (~300ms).
  • Result: Driver hears the strategy clearly via In-Ear Monitors (IEMs).

AI Chief Audio: "One-point-two-six seconds to P-one. You're closing. Carry more speed through turn seven. Tires are good for five more laps."

API Integration Points

1. Device → API: Send Driver Query

// AI Chief Radio device sends:
{
  "deviceId": "ai-chief-55",
  "carNumber": 55,
  "lapNumber": 18,
  "driverQuery": "what's my gap to the leader",
  "timestamp": "2025-11-18T20:15:03Z"
}

// API responds:
{
  "aiResponse": "One-point-two-six seconds to P-one...",
  "telemetrySnapshot": { /* current data */ },
  "priority": "high"
}

2. API → Device: Real-Time Updates

// API pushes proactive alerts: { "type": "proactive-alert", "carNumber": 55, "message": "Tire wear reaching 75%. Consider pit stop in 3 laps.", "urgency": "high", "ttsAudio": "base64-encoded-audio-data" }

// Device plays audio immediately


### 3. Device → API: Telemetry Upload

```javascript
// Device continuously sends telemetry:
{
  "carNumber": 55,
  "timestamp": "2025-11-18T20:15:03Z",
  "biometrics": {
    "heartRate": 158,
    "hrv": 42,
    "skinTemp": 35.2
  },
  "audioMetrics": {
    "micQuality": 94,
    "noiseCancellationActive": true
  }
}

🎬 Complete User Flow Example

Scenario: Driver asks about tire wear

1. Driver speaks (in car, ~110dB engine noise)

Driver: [Presses PTT button]
"Hey Chief, how are my tires?"

2. Device processes (AI Chief Radio hardware)

• Bone conduction mic captures voice (50ms)
• Noise cancellation removes engine (20ms)
• Whisper STT: "how are my tires" (300ms)
• Send to API via 5G (30ms)

3. Your API processes (Agent2Agent backend)

{
  "carNumber": 55,
  "lapNumber": 18,
  "driverQuery": "how are my tires"
}

→ Car Telemetry Agent: Load car #55 data
→ Engine/Tire Agent: Calculate wear = 67%
→ Weather Agent: Track temp stable
→ LLM Insight Agent: Generate response
→ Return: "Tires at 67%. Good for 5 more laps..."

4. Device responds (AI Chief Radio hardware)

• Receive response text (100ms)
• Azure TTS synthesis (300ms)
• Stream audio to IEMs (100ms)

5. Driver hears (~2 seconds after speaking)

AI Chief: 🔊
"Tires at sixty-seven percent. You've got good life left. 
Track temperature is stable. You can push for another 
five laps before considering conservation mode."

6. Driver acknowledges

Driver: [Quick button press = ACK]
Device: [Green LED flash]

📊 Data Flow Summary

┌─────────────────────────────────────────────────────────────┐
│  WHAT FLOWS WHERE                                           │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  Device → API:                                              │
│  • Driver voice queries (text after STT)                    │
│  • Device status (battery, connection quality)              │
│  • Button presses / interactions                            │
│                                                             │
│  API → Device:                                              │
│  • AI coaching responses (text for TTS)                     │
│  • Proactive alerts (without driver asking)                 │
│  • Telemetry updates (position, gaps, lap times)            │
│  • System notifications (warnings, errors)                  │
│                                                             │
│  API → LLM (OpenAI/Claude):                                 │
│  • Race context from 8 agents                               │
│  • Driver query + telemetry snapshot                        │
│  • Historical conversation context                          │
│                                                             │
│  LLM → API:                                                 │
│  • Natural language coaching response                       │
│  • Urgency level (low/medium/high/critical)                 │
│  • Suggested follow-up questions                            │
│                                                             │
└─────────────────────────────────────────────────────────────┘

This is a complete, integrated system that showcases:

  • Real hardware innovation (the device)
  • Advanced AI/ML (the 8 agents)
  • Practical application (racing)
  • Market opportunity ($850M)
  • Technical execution (code + detailed specs)

Challenges we ran into

Integrating multiple data streams and simulating a real time race strategy for the purpose of a demo

Designing coaching prompts that are fast, actionable, and never distracting

Ensuring the system performs with responsiveness

Working with LLM models to create a realistic looking design for the helmet and AI device integration

Accomplishments that we're proud of

Delivering tactical coaching responses in under two seconds

Creating a fully autonomous AI race engineering stack

Developing a holistic driver-and-machine performance model

Design of seamless communication between the helmet device and HeartBeat dashboard

What we learned

We learned that the right AI insights at the right moment can dramatically elevate a driver’s pace, confidence, and safety. We also discovered the importance of intuitive audio coaching and the power of combining physiological, mechanical, and environmental data into one unified system.

What's next for X-RaceIQ

We’re expanding into more racing disciplines, refining predictive models, and developing deeper team-side analytics.

Next up: enhanced driver biometrics and biometric tracking models, more adaptive coaching personalities, and expanded integrations to become the ultimate AI teammate in motorsport we are proud of. Additionally, safeguards to allow crew members to override AI suggestions before relay to drivers to prevent hallucinations in LLM models affecting real life driving performance.

Built With

Share this project:

Updates