Vendira — AI Voice & Omnichannel Call Center Platform
Multi-tenant AI voice platform deploying real-time, human-like voice agents for customer engagement and support automation.
Role
Full Stack AI/ML Developer
Timeline
Jan 2026 — Present
Stack
Overview
Vendira is a multi-tenant AI-powered call center and omnichannel automation platform that deploys real-time, human-like AI voice agents for customer engagement, lead handling, and support automation. The platform automates inbound and outbound phone calls using a full AI voice stack — Deepgram for real-time speech-to-text, ElevenLabs for natural text-to-speech synthesis, and Groq + OpenAI for ultra-fast LLM inference — all orchestrated over Twilio's telephony infrastructure. Businesses configure AI agents, monitor live sessions, and manage multi-channel communication across voice, SMS, and chat from a unified React dashboard.
Architecture & Decisions
Real-Time AI Voice Pipeline
Architected a full-stack AI voice pipeline combining Deepgram's real-time speech-to-text, Groq and OpenAI for sub-second LLM inference, and ElevenLabs for natural voice synthesis. Phone calls are orchestrated via Twilio, enabling inbound and outbound AI-driven conversations with human-like response latency. The system handles intelligent call routing, context-aware multi-turn dialogue, conversation memory, sentiment detection, and graceful handoff to human agents when needed.
Multi-Tenant SaaS Dashboard & Omnichannel Automation
Built with React, Shadcn UI, and Tailwind CSS, the management dashboard allows businesses to configure AI agent personas, define call flows and scripts, monitor live sessions in real time, and analyse conversation analytics. Supports omnichannel automation across voice calls, SMS, and chat through Twilio's unified communication API. Multi-tenant architecture isolates each business's agents and data. FastAPI backend handles high-throughput real-time events via WebSockets, with PWA support for mobile access.
The Hard Part
Callers notice latency immediately. The challenge was chaining four different AI services — speech-to-text, LLM inference, voice synthesis, and telephony — into a single pipeline that still feels like a real conversation, not a queue of API calls.
Outcomes
Latency
Sub-second
Model
Multi-tenant
- —Full AI voice stack (Deepgram, Groq, OpenAI, ElevenLabs) orchestrated over Twilio
- —Intelligent call routing, multi-turn dialogue, and sentiment-aware human handoff
- —Unified dashboard for configuring agents and monitoring live sessions across voice, SMS, and chat
- —PWA support for mobile access to the management dashboard
Up Next
View all
Regulatory Compliance Platform
Enterprise regulatory reporting platform for financial institutions — 80% faster reporting, 65% better data accuracy.
Read case study