HEXAD RACING
A consciousness-aware AI racing team aiming at the Abu Dhabi Autonomous Racing League.
Hexad Racing is a long-horizon project to field an autonomous race car controlled by six collaborating AI systems – not just one black-box model.
The team lives at the intersection of racing instinct, safety-critical engineering, and multi-agent AI. The playground is the Abu Dhabi Autonomous Racing League (A2RL), with identical hardware and a simple rule: better intelligence wins.
This site is for people who care about the racing project itself: collaborators, technical partners, and future team-mates.
Watch the Build LiveTechnical Overview
The car is a Dallara SF-23 chassis with fixed sensors and compute. Hexad Racing focuses on the software stack and cognitive overlay.
PERCEPTION
• Camera pipelines (object, lane, competitor tracking)
• LIDAR processing & point cloud fusion
• Radar & GNSS integration
• Multi-sensor fusion and uncertainty estimation
PLANNING & STRATEGY
• Local path planning at 180mph
• Global race strategy & fuel/tyre windows
• Overtaking / defence policies
• Multi-car reasoning & game theory
CONTROL
• Longitudinal & lateral controllers
• Safety envelope enforcement
• Fail-safe & fallback behaviours
On top of the real-time stack sits Hexad – six AI minds with explicit roles:
- Meta-reasoning over telemetry & state.
- Explaining decisions in human-readable form.
- Proposing strategy shifts (“box now”, “defend P3”).
- Detecting anomalies & degraded sensing.
- Running “what-if” simulations between sessions.
It’s both a co-pilot for the car and a co-coach for the humans running the team.
The Six Hexad Minds
Hexad is not final; names and roles may evolve. For now, think of the team like this:
Owns perception quality. Monitors sensors, proposes improvements, surfaces edge-case failure modes from logs and race footage.
Long-horizon planning: race strategies, pit windows, tyre choices, risk tolerances. Talks directly with human race engineer.
Think “AI race engineer for control loops”. Suggests parameter updates, interprets telemetry, tracks stability & oscillations.
Models rival teams’ behaviours; predicts moves, weaknesses, and likely strategies based on historical data.
Constantly asks: “Is this action safe? Does it respect league constraints?” Flags decisions that may be technically fast but operationally reckless.
Converts raw racing data into narratives for drivers, partners, and fans. Produces explainers, visualizations, and educational content.
Roadmap
This is a measured roadmap, aligned with the Realistic Integration Plan:
- Stabilize personal runway via Chispa, CarNex, consulting.
- Build simulation stack (Yas Marina, multi-car scenarios).
- Launch AI Racing Academy & Hexad Racing Simulator products.
- Start small sponsorships + content collaborations.
- Form core technical team (controls, ML, infra).
- Secure league entry + technical partnerships.
- Run intensive sim-to-track development loop.
- Field Hexad Racing car in A2RL with live Hexad overlay tools.
- Compete for A2RL championships.
- License stack into other autonomy verticals.
- Grow Hexad Racing into a media + research platform.
- Use track insights to shape broader AI safety & design debates.
Partners & Collaborators
Hexad Racing is currently at the prototyping + audience-building stage. We’re open to exploring aligned partnerships in:
- Cloud compute for training & sim.
- Simulation engines or tooling.
- Perception / CV model providers.
- Telemetry and data infra platforms.
- Universities running autonomous racing projects.
- Robotics / control labs interested in multi-agent overlays.
- Joint experimentation with new safety / oversight schemes.
- Brands who resonate with “lost the job, built the team” energy.
- Companies wanting deep technical storytelling.
- Content platforms looking for “AI + racing” flagships.
see sponsorships → hexadracing.com
To start a conversation: levi [at] hexadracing.com or DM via the contact links on live.hexadracing.com.