Back to Work

/ Case.Study

01
/Apr 2026 – Present

AxonVoice AI

Real-Time AI Voice Assistant & Automated Booking System

A self-hostable, multi-tenant real-time voice agent platform supporting Sinhala, Tamil, and English. Four isolated microservices behind a YARP API Gateway with JWT auth, Redis rate limiting, and WebSocket upgrades.

Project ID

axonvoice

Timeline

Apr 2026 – Present

Stack Footprint

8 Technologies

Delivery Tracks

4 Key Areas

AxonVoice AI project case study preview

/ Project.Overview

What Was Delivered

A self-hostable, multi-tenant real-time voice agent platform supporting Sinhala, Tamil, and English. Four isolated microservices behind a YARP API Gateway with JWT auth, Redis rate limiting, and WebSocket upgrades.

/ Delivery.Highlights

Implementation Notes

01

Zero-copy PCM audio streaming via System.IO.Pipelines as bidirectional WebSocket proxy to Gemini Live API

02

Semantic Kernel integration intercepting LLM function_call events mid-stream with parallel DB queries

03

Cost-optimised RAG pipeline using Qdrant + local Ollama, reducing prompt token sizes by over 90%

04

Concurrency-safe dual-table booking system with SQL CTEs preventing race conditions