⚡ The Elevator Pitch
Autex is an autonomous automotive OS bridging hardware telemetry and human reasoning. Our hybrid edge-cloud architecture connects directly to OBD-II hardware via Web Serial. A local "Sentinel" AI detects issues offline (0ms latency), while a cloud "Strategist" (Gemini 2.5 Flash) formulates repair plans. We issue digital "Proof of Condition" certificates to boost resale value and provide ROI-driven maintenance insights, solving information asymmetry in the $4T automotive economy.
💡 Inspiration
We were inspired by the "Check Engine Light Anxiety." Modern vehicles are computers on wheels, yet most owners are disconnected from their hardware by cryptic codes. We wanted to build an agent that doesn't just "report" data but "reasons" through it—bridging the gap between a car's mechanical heartbeat and a human's need for strategy, trust, and transparency.
🛠 What it does
Autex is a Progressive Web App (PWA) that functions as an Autonomous Automotive Operating System:
- The Neural Handshake: A dual-agent system where a local "Sentinel" handles high-speed perception and a cloud "Strategist" handles high-level reasoning.
- Animal Temperament Cores: Specialized reasoning modes—Stallion (Performance), Nomad (Endurance), and Workhorse (Audit)—that shift the AI's diagnostic focus in real-time.
- Web Serial Hardware Link: Connects directly to OBD-II hardware from the browser with zero drivers.
- Proof of Condition: Issues verified, AI-audited health certificates to increase vehicle resale value by up to 5%.
- Bilingual Voice Briefings: Real-time diagnostic summaries in English and Arabic using Gemini TTS.
🏗 How we built it
- AI Core: Built with Genkit and Gemini 2.5 Flash for agentic reasoning and tool-calling.
- Framework: Next.js 15 (App Router) for a high-performance, PWA-first experience.
- Database: Firestore with a specialized 40MB Edge Cache configuration for "Offline Survival" in garages.
- Connectivity: Web Serial API for direct hardware-to-browser telemetry streaming.
- UI/UX: ShadCN and Tailwind CSS, optimized for both LTR (English) and RTL (Arabic) display.
🚧 Challenges we ran into
- Telemetry Latency: Managing high-frequency sensor data without blocking the UI thread. We solved this with a buffered sampling strategy for the Edge Sentinel.
- Cloud Reliability: Ensuring the AI doesn't hallucinate in signal-dead zones. We implemented "Limit-Awareness" where the agent identifies when it's offline and reverts to local perception models.
- Quota Management: Handling the Gemini Free Tier limits. we built a "Neural Cooldown" timer and exponential backoff logic into our AI flows.
🏆 Accomplishments that we're proud of
- Zero-Friction Hardware Link: We achieved a direct serial connection to the car without requiring any app store downloads or Python drivers.
- The Animal Modes: Creating a UI where the AI's "Temperament" actually changes the technical logic of the diagnosis.
- Bilingual Excellence: A fully functional, voice-enabled Arabic and English diagnostic agent.
🧠 What we learned
- The "Fast and Slow" Handshake: How to balance "fast" local perception (Sentinel) with "slow" cloud reasoning (Strategist) to create a reliable agentic system.
- Agentic Reliability: The importance of giving AI agents the ability to say "I don't know" or "I need more data" to maintain technical authority.
🚀 What's next for AUTEX
- Live Parts Marketplace: Integration with global parts APIs for automated one-tap ordering of diagnosed components.
- Predictive Maintenance for Fleets: Scaling the OS to manage hundreds of connected assets in real-time.
- Hardware Partnerships: Developing a proprietary Autex OBD-II "Green Box" for even higher-fidelity telemetry.
🖥 Dashboard Components in Detail
The Autex Command Center is designed for high-fidelity technical oversight:
- KPI Command Bar: Real-time tracking of the Health Performance Index (HPI), RPM, and System Temperature. It features an Edge Analysis counter showing live inference runs performed locally by the Sentinel.
- Spectral Telemetry Chart: A dual-view interface allowing users to toggle between Time Domain (amplitude) and Spectrogram (frequency) analysis, essential for identifying subtle engine harmonic imbalances.
- Agent Execution Log: A real-time terminal showing the "thinking" process of the AI agents. It highlights CAR-bench Fast-Sync events and provides reliability scores for every decision.
- Active Incident Log: A prioritized feed of anomalies. Each alert includes an AI-generated Maintenance Strategy, part availability from the MongoDB warehouse, and integrated one-tap commerce buttons.
- Maintenance Insights: A dedicated ROI dashboard that calculates cost avoided via early detection and monitors Predictive Energy Efficiency to detect power drift.
- Proof of Condition Module: An auditor interface for generating digital certificates. It uses AI to verify telemetry consistency and estimate the positive impact on the vehicle's resale value.
- Autex AI Mechanic (Chat): A persistent, floating agent that maintains context of the live sensor stream, allowing users to ask questions like "What is that rattling sound?" or "Is it safe to drive 500km?"
Note:
I have difficulty writing summaries. Artificial intelligence helped me write this summary.
I conceived the idea, drawing inspiration from nature, and I created the architecture and design. Artificial intelligence assisted me with the code.
Log in or sign up for Devpost to join the conversation.