Case Study: CTD AI - AI Growth Platform for Dental Clinics

Project at a Glance

Role: Full-Stack / AI Platform Engineer
Duration: Feb 2026 - Present
Live: ai.connectthedoc.com
Stack: React, Vite, TypeScript, Node.js, Express, Prisma, PostgreSQL, Redis, BullMQ, OpenAI/OpenRouter, Google Places, BrightLocal, DataForSEO, WordPress, Google Business Profile, AWS S3, Docker, Nginx
Scale: ~280,000 lines of code, 54 Prisma models, background worker pipeline + product UI
Team: Product engineering for Connect The Doc - I built and extended the AI growth system end to end

Problem

Dental clinics know they need local SEO, Google Business activity, and consistent content, but they do not have a marketing team. Existing tools give them dashboards or disconnected AI writers. CTD AI turns clinic onboarding into an automated growth engine: discover the practice, map competitors, generate on-brand content, publish it, and keep ranking/content intelligence fresh.

What I Built

Guided clinic onboarding that creates real data, not empty accounts

Google Places-powered clinic search and confirmation
AI practice summary + service guesses, neighbourhood targeting, and keyword selection with volume/difficulty signals
Competitor discovery and first content/savings report so value shows up before the clinic "figures out the product"

Competitor intelligence

Auto-discover nearby competitor clinics
Local map-pack rank checks across target neighbourhoods and keywords (BrightLocal)
Competitor GBP/blog activity tracking with weekly refresh jobs so rankings do not go stale

Content pipeline (plan -> draft -> queue -> review -> publish)

Topic planning from services, keywords, and competitor gaps
AI drafting for blog posts and Google Business posts in the clinic's voice (Practice DNA)
Scheduled review queue so content ships steadily instead of dumping 20 drafts at once
Publish to WordPress and Google Business Profile via real integrations

AI assistant grounded in live clinic data

Chat that answers using clinic context, rankings, and content state - not a generic chatbot wrapper
Practice DNA brand-voice profile so generation stays consistent across posts and chat

Platform engineering

Express/TypeScript API with Prisma + Postgres
BullMQ workers for enrichment, competitor scans, content generation, crawling, and scheduled jobs
Docker Compose staging/production with Nginx TLS on ai.connectthedoc.com
Production DB sync tooling so local debugging can use real-shaped data safely

Architecture Highlights

Clinic UI (React + Vite)
        |
        v
 Nginx TLS  --->  API (Express + Prisma + Postgres)
                        |
                        +-- Onboarding / Places / Keywords
                        +-- Competitor intelligence
                        +-- Content + Practice DNA
                        +-- WordPress / GBP integrations
                        |
                        v
                 Redis + BullMQ workers
                        |
                        +-- Rank scans / SEO crawls
                        +-- Content generation jobs
                        +-- Publish + refresh schedules

Hard Problems I Solved

Onboarding that must feel magic and still be correct: Places confirmation, service inference, neighbourhoods, and keyword metrics have to compose into a coherent clinic graph on day one.
AI that is operational, not theatrical: Drafts enter a queue with review and publish paths. Nothing "AI-generated" ships without clinic/admin control.
Local SEO data is multi-vendor: BrightLocal, DataForSEO/SerpApi, Places, WordPress, GBP. Designed service boundaries and workers so failures in one provider do not take down the product.
Brand voice at scale: Practice DNA keeps hundreds of posts on-brand without rewriting prompts by hand for every clinic.

Results & Impact

Production system live at ai.connectthedoc.com
Clinics go from login -> discovered practice -> competitor map -> first content batch in one guided flow
Recurring weekly intelligence + content cadence without hiring a marketing ops person
Large, maintainable codebase with clear product docs for every major feature area

What This Shows About Me

I build AI products that plug into real business workflows (SEO, GBP, WordPress), not toy chat demos
I can design long-running worker systems and multi-API orchestration
I care about operator UX: queues, review, scheduling, and trust controls around generation
I document and ship production platforms at serious scale (~280k LOC)