DEV Community

Jon Lyle
Jon Lyle

Posted on

I built a hands-free AI mechanic. Here's what I learned about voice diagnostics at scale.

I'm Jon — third-generation mechanic out of Detroit. Last week, I shipped something that's been keeping me up at night: Pocket Mechanic — a hands-free AI diagnostic system that actually listens.

Not "upload a file and wait." Not "click buttons and read code definitions." Talk to it. It talks back. It understands.

The Problem I Was Trying to Solve

You're under a car. Hands covered in grease. You pull a code. You need diagnostics NOW. But your hands are dirty, your phone's in your pocket, and you're not about to stop what you're doing to type.

So you either:

  • Memorize the code and Google it later (lose context)
  • Yell at your buddy to look it up (interrupts their work)
  • Stop working, clean your hands, use your phone (wastes 5-10 minutes per code)

That's the gap. That's what I built for.

How Pocket Mechanic Works

Full-duplex voice conversation. You talk, it listens in real-time. No "wait for the beep" nonsense. No lag. It's like talking to another tech in the bay.

  • Upload your datalog once (FORScan, HP Tuners, Autel, OBD2 CSV)
  • Start talking: "What's causing the misfire?"
  • It analyzes the data in context of what you're saying
  • Gives you root causes, ranked by severity
  • Remembers your previous diagnostics (knows your customer's truck history)
  • Knows localization (salt-belt fleet truck vs. highway commuter)

The Tech Stack (For the Curious)

  • Voice input: Real-time transcription, no latency
  • LLM backbone: Automotive-specific training (not generic ChatGPT)
  • Data analysis: CAN bus visualization, freeze frame parsing, pattern recognition
  • Conversational memory: Persists context across sessions
  • Output: Text, voice, or shareable diagnostic reports

Current Status

Free through Memorial Day week. Full Pro access. No credit card. No auto-charge. Just talk, get answers, subscribe when you're ready.

Post-beta pricing: $49/mo basic, $99/mo pro, $199/mo team

URL: https://www.grease-ai.com

What I Learned Building This

  1. Mechanics don't want to learn new tools. They want tools that fit into their existing workflow. Hands-free is the only way.
  2. Voice is the interface. Not buttons, not forms, not "contact sales." Just talk.
  3. Data context matters more than code definitions. Every tech can Google a P0300. What they need is "here's what YOUR data is telling me."
  4. The blue-collar SaaS market is wide open. And it's hungry for tools built by people who actually use them.

This is Grease AI's flagship feature. If you turn wrenches for a living, this was built for you.

Top comments (0)