Nadeem Akhtar
01 Home 02 My Journey 03 Products Engineered Active 04 Onboarding Workspace
Download Resume PDF
© 2026 Nadeem Akhtar
Back to Projects
SaaS Platform Core

Voixana AI Receptionist Platform

Multi-tenant AI receptionist SaaS platform for businesses supporting mobile, web, and admin interfaces.

Flutter Web/MobileDart 3.9GetXFirebase CoreCloud FirestoreCloud FunctionsFlutter ScreenUtilFirebase Messaging

Role

Lead Architect

Duration

6 Months (SaaS Core Built)

Platform

Production App

Status

Codebase Complete

Demo Not Available
Voixana AI Receptionist Platform Banner
Project Overview

System & Product Summary

Voixana is a multi-tenant AI receptionist SaaS platform engineered for small-to-medium businesses. Built with Flutter Web/Mobile and powered by Cloud Functions and Firebase, the platform integrates AI model routing, Twilio voice streams, and WebSocket audio channels to automate customer phone inquiry handling, appointment scheduling, and Stripe billing.

The Challenge

Core Problem & Friction

Small businesses lose critical customer inquiries during peak hours or after-hours, leading to lost revenue, while hiring full-time human receptionists introduces steep operational overhead.

Systemic Constraints

Multi-tenant architecture separation, Stripe automated subscription checkout flows, sub-300ms transcription latencies, and high concurrency call queues.

Feasibility & Discovery

Engineering Research

I researched WebSocket communication protocols and LLM latency profiles. I analyzed concurrent transcription buffers and Twilio voice stream bindings to ensure zero audio packet dropouts during active call transfers.

System Design

Software Architecture

Frontend Flutter Client (Dart 3.9 & GetX)
Backend API Cloud Functions & NodeJS API
AI Engine OpenAI Call Agent Routing
Database Cloud Firestore
Auth / Security Firebase Auth
Deployment Internal Architecture
Intelligence Layer

AI & Machine Learning Architecture

Engineered dedicated machine intelligence pipelines utilizing OpenAI Call Agent Routing for inference and dynamic routing.

The Solution

Solution Implementation

We engineered a real-time voice receptionist framework that buffers live Twilio call streams via WebSockets into Redis arrays and routes transcripts through OpenAI GPT endpoints to execute automated inquiry responses and booking actions in sub-300ms turns.

Tradeoffs & Rejected Approaches

I rejected Twilio's default recording APIs in favor of live WebSocket audio streaming. Although WebSocket links require persistent server scaling, they allow real-time transcriptions and sub-200ms model responses.

Engineering Optimizations

I optimized the transcription pipelines by buffering audio packets in Redis memory arrays, preventing database write blockages during peak hours.

Metrics & Results

Quantifiable Impact & Metrics

Automated 92% of routine business calls, integrated recurring payments, and secured 100% data partition isolation across tenant databases.

Technology

Technology Stack

Flutter Web/MobileDart 3.9GetXFirebase CoreCloud FirestoreCloud FunctionsFlutter ScreenUtilFirebase Messaging
Capabilities

Key Features & Functional Scope

Multi-tenant SaaS dashboard managing call history and transcriptions

AI-powered semantic routing and call forwarding configurations

Subscription and automated checkout flows built via Stripe integrations

Visuals

Project Showcase

Project screenshot
Takeaways

Lessons Learned & Takeaways

WebSockets require strict heartbeat monitoring routines to recover connection paths after mobile network switches.

Roadmap

Future Scope & Improvements

Integrating local voice synthesis models to reduce reliance on external TTS API calls.

Ready to Build Something Extraordinary?

Initiate the onboarding workspace to scope out architecture requirements, timelines, and budget parameters.

Initiate Onboarding Workspace →