Available for work·Book a call
All Projects/Dentva.ai — AI Voice Receptionist

Full Stack AI Developer

Dentva.ai — AI Voice Receptionist

24/7 AI voice receptionist for dental clinics — answers patient calls, books appointments, verifies identities, and handles FAQs with HIPAA-focused architecture

Next.jsTailwind CSSshadcn/uiPython FastAPILangChainLangGraphAzure OpenAIAWS Transcribe MedicalAWS PollyAurora PostgreSQL + pgvectorRedisAmazon CognitoiPlum / Telnyx (Telephony)AWS SNS / SESAmazon S3AWS ECS FargateCloudWatch
Dentva.ai — AI Voice Receptionist

Overview

Dentva.ai is a HIPAA-focused AI voice automation platform that replaces traditional dental clinic receptionists with a conversational AI that never sleeps. Dental practices lose patients every day to missed calls and limited front-desk hours. Dentva.ai answers every inbound call, understands natural language requests, books or reschedules appointments in real time via PMS integration, verifies patient identity via OTP before sensitive actions, and answers clinic-specific FAQs using semantic vector search — all while keeping protected health information secure through encryption, audit logging, and role-based access control.

The Challenge

Building a voice AI that works reliably in a healthcare context required solving three hard problems simultaneously: sub-second latency for natural conversation feel (patients hang up if the AI sounds laggy), strict HIPAA-focused data handling for every transcript and patient record, and reliable PMS integration with graceful fallback logic when the practice management system is temporarily unavailable. On top of that, each dental clinic has its own services, providers, insurance policies, and hours — requiring a multi-tenant architecture where every clinic's knowledge base is fully isolated.

How We Solved It

1

Real-Time Voice Processing Layer

Built an end-to-end voice pipeline integrating telephony (iPlum / Telnyx), AWS Transcribe Medical for speech-to-text, Azure OpenAI for response generation, and AWS Polly for natural voice synthesis. Streaming architecture keeps latency low enough that conversations feel natural rather than robotic.

2

AI Conversation Engine with LangChain + LangGraph

The AI orchestration layer uses LangChain and LangGraph to manage multi-step dialogue workflows: speech recognition → intent detection → context tracking → patient verification → PMS lookup or knowledge retrieval → response generation → voice synthesis. LangGraph's stateful graph architecture ensures conversations remain coherent across multi-turn interactions.

3

PMS Integration & Appointment Automation

Integrated with dental Practice Management Systems to handle real-time availability checks, new bookings, rescheduling, and cancellations — including provider and appointment type selection. A fallback scheduling layer ensures zero missed appointments even during PMS downtime.

4

OTP-Based Patient Verification

Sensitive appointment actions (modification, cancellation) are protected by a caller identification + OTP verification flow using AWS SNS. Only verified patients can modify their own records, preventing unauthorized access to PHI.

5

Semantic Knowledge Base with pgvector

Each clinic maintains an isolated vector knowledge repository stored in Aurora PostgreSQL with the pgvector extension. Clinic staff upload FAQ documents or add Q&A pairs directly through the dashboard. During calls, the AI performs semantic similarity search to retrieve the most relevant answers for patient queries about services, hours, insurance, and treatments.

6

Clinic Admin Dashboard

A Next.js + shadcn/ui dashboard gives clinic admins full visibility and control: call history with duration/outcome, HIPAA-redacted transcripts with AI summaries, PMS data sync (providers, appointment types), knowledge base management, and call analytics. Authentication is secured through Amazon Cognito with role-based access controls.

7

HIPAA-Focused AWS Infrastructure

The entire platform runs on AWS ECS Fargate with encrypted Aurora PostgreSQL databases, S3 for secure storage, CloudWatch for monitoring and audit logging, and end-to-end PHI redaction in all exported transcripts. Every layer is designed around the principle of minimum necessary access.

Admin Dashboard

Results & Impact

24/7 Patient Call Coverage

Every inbound patient call answered — no missed calls during lunch, after hours, or on weekends.

Automated Appointment Booking

Patients book, reschedule, and cancel appointments entirely through the AI voice interface, with real-time PMS sync.

OTP-Verified Identity Protection

All sensitive appointment actions gated behind caller ID + one-time password verification before any PHI is accessed or modified.

Semantic FAQ Resolution

pgvector-powered semantic search retrieves the correct clinic-specific answer across services, insurance, hours, and treatments — even for paraphrased or ambiguous patient questions.

HIPAA-Focused Transcripts

All call transcripts are stored with PHI redaction, audit logging, and role-based access — exportable by clinic admins from the secure dashboard.

Multi-Tenant SaaS Architecture

Each dental clinic operates in a fully isolated data environment, enabling safe SaaS expansion to multiple practices on shared infrastructure.

Technical Architecture

FrontendNext.js, Tailwind CSS, shadcn/ui — clinic admin dashboard
BackendPython FastAPI — voice pipeline, appointment logic, verification, knowledge retrieval
AI / LLMAzure OpenAI (GPT-4), LangChain + LangGraph — multi-step voice conversation orchestration
VoiceAWS Transcribe Medical (STT), AWS Polly (TTS), iPlum / Telnyx (telephony)
DatabaseAurora PostgreSQL + pgvector (semantic FAQ search), Redis (session cache)
AuthAmazon Cognito — role-based access control for clinic staff
MessagingAWS SNS (OTP delivery), AWS SES (email notifications)
StorageAmazon S3 — encrypted transcript and document storage
InfraAWS ECS Fargate — containerized deployment with CloudWatch monitoring and audit logging

Have a similar project?

I'm open to freelance projects and full-time roles. Let's build something great together.