TechLens
Your AI Diagnostic Copilot
Hands-free voice + camera diagnosis for auto technicians. One conversation. Three professional documents.
See It In Action
TechLens
Hands-free voice + camera diagnosis for auto technicians. One conversation. Three professional documents.
See It In Action
Techs are shoulder-deep in engine bays. They can't stop to type. Diagnostic notes end up illegible, incomplete, or missing entirely.
Service advisors manually translate tech-speak into something customers can understand. Details get lost, calls get confused, trust erodes.
Manufacturer reps bounce warranty claims for incomplete documentation. Missing measurements, no TSB references, no supporting data.
Every bounced warranty claim costs a shop $400–800 in lost labor.
Every DTC code, every measurement, every diagnostic step — structured, searchable, ready for the repair order. Part numbers included. TSB references linked. No more deciphering handwritten shorthand.
Plain English your customer actually understands. No jargon, no acronyms, no 'valve body assembly.' Empathetic tone, clear next steps, transparent on timeline. Ready to hand over at pickup or email directly.
The complete diagnostic narrative a manufacturer field rep needs to approve warranty coverage. Full measurement history, every DTC logged in sequence, TSB cross-references, NHTSA complaint volume. Built to survive scrutiny.
Every DTC code, every measurement, every diagnostic step — structured, searchable, ready for the repair order. Part numbers included. TSB references linked. No more deciphering handwritten shorthand.
Plain English your customer actually understands. No jargon, no acronyms, no 'valve body assembly.' Empathetic tone, clear next steps, transparent on timeline. Ready to hand over at pickup or email directly.
The complete diagnostic narrative a manufacturer field rep needs to approve warranty coverage. Full measurement history, every DTC logged in sequence, TSB cross-references, NHTSA complaint volume. Built to survive scrutiny.
Real output from a single TechLens session — generated in under 3 seconds
Enter year, make, model, trim, and the customer's complaint. Gemmy loads the relevant knowledge base and starts listening immediately. Always on — just talk.
Speak naturally. Fumble your words. Change your mind mid-sentence. Gemmy handles real conversation, not rigid commands.
Gemmy pulls from 21 TSBs, 3 recalls, and 175 NHTSA complaints. 80% of questions answered instantly from pre-loaded context — no tool call latency.
Activate the camera with your voice. Point your phone at any component — Gemmy sees what you see and identifies parts, damage, and next steps. Say "Stop looking" to turn the camera back off.
Barge in mid-sentence. Gemmy stops immediately and pivots. This is a conversation, not a queue.
Three professional documents generate automatically. No typing. No copy-paste. No "what did the tech mean by chk tsb??"
Voice + camera input — hands-free, under the car
WebSocket + MediaStream capture
React 19 + Tailwind v4Orchestrates the four-agent pipeline
us-central1 · Python + FastAPIPre-filters 70KB knowledge base → 2KB vehicle-specific context. Injects into Live Agent system prompt.
gemini-2.5-flash · runs onceReal-time voice + vision conversation. Barge-in supported. Hands-free.
gemini-2.0-flash-live · ADKLive synthesis visible to the tech. TSB matches, findings, severity ratings update in real time.
Visible in UI · Updates continuouslyBackground async analysis. TSB matching, NHTSA cross-reference, severity ratings.
gemini-2.5-flash · async21 TSBs · 3 Recalls · 175 NHTSA complaints
Tool calls ↔ Structured JSONTakes full transcript + synthesized findings → generates three professional documents
gemini-2.5-flashDTCs · Part numbers · Procedures
Plain English · No jargon
Full diagnostics · History
Richard Adair spent seven years in the Subaru service department—starting as a lube tech doing oil changes in the express lane, moving to service writing, then main shop work, and finally to lead dispatcher and management. Lube tech. Service writer. Lead dispatcher. He watched this problem from every seat in the shop. TechLens was built by someone who lived every role—not someone who read about it.
Scan the QR code to open TechLens on your phone — or tap the link below on desktop. No install needed. Works in any browser with mic + camera access.
Launch TechLensShop owner, service advisor, or technician? Be one of the first to use TechLens in a real bay. Drop your email and we'll reach out when we're ready for testers.
Interested in investing, integrating TechLens into your platform, or exploring a partnership? Leave your email and tell us what you're thinking.
Ready to move? Skip the form. Reach out directly — I respond to every message.
TechLens was born in a Subaru dealership service department. The founder, Richard Adair, spent seven years working every role in the shop — lube tech, service writer, main shop technician, lead dispatcher, and management. Every day, he watched the same documentation failures cost shops time, money, and customer trust. Techs know what’s wrong with the car but can’t capture it fast enough. Advisors spend twenty minutes translating. Manufacturer escalations get bounced for missing data. Three audiences need the same information written three different ways, and nobody has time to do it.
TechLens runs a multi-agent AI pipeline on Google Cloud. When a tech starts a session, the Intake Agent compresses a 70KB knowledge base — covering TSBs, recalls, and NHTSA complaints — down to roughly 2KB of vehicle-specific context. That’s a 35:1 compression ratio, injected directly into the Live Agent’s system prompt before the conversation begins.
The Live Agent uses Google’s Gemini Live API for real-time bidirectional voice and vision streaming. The tech talks naturally and can interrupt mid-sentence. A Synthesizer Agent watches the transcript in the background, matching symptoms to TSBs and feeding enriched context back into the conversation — a continuous loop where the AI gets smarter as the session progresses.
When the session ends, the Writer Agent generates three professional documents automatically: structured Tech Notes for the repair order, a plain-English Customer Summary, and a formal Escalation Brief built to survive manufacturer scrutiny.
Pre-filtering beats retrieval. Injecting context at session start is faster and more reliable than runtime RAG for voice conversations where latency kills the experience. 80% of questions are answered from pre-loaded context with zero tool calls — no dead air.
Building for people who can’t touch a screen forces fundamentally different UX decisions. Every interaction must be voice-first and camera-optional. Domain expertise turned out to be the real differentiator — knowing what a tech needs on a repair order vs. what a customer needs to hear vs. what a manufacturer rep requires can’t be Googled.
Replace the phone camera with wearable optics for true hands-free diagnosis.
Write directly to CDK, Reynolds & Reynolds, and Tekion so tech notes land in the RO automatically.
Expand beyond Subaru to Toyota, Honda, Ford, GM, and more.
Aggregate session data to surface patterns — which symptoms correlate with which fixes across thousands of repairs.
Gemini Live API · Google ADK · Cloud Run · Cloud Firestore · React 19 · Tailwind v4 · FastAPI · Python · WebSocket