How I built Lytespeed — an AI voice platform for 22 Indic languages — from the ground up
In May 2024 we started building. By September 2024, Lytespeed was live — a full AI agent platform with a visual workflow builder, 22 Indic language support, multi-channel deployment, and a working SaaS pricing model. Four months, one lead engineer, zero existing codebase.
22
Indic language voice support
6 mo
Zero to live SaaS product
Client:
Etheral Covenant
Duration:
Apr 2022 – Oct 2024
Location:
Bangaluru, India
Role:
Lead Product Engineer → Engineering Manager
The Problem
Businesses wanting to deploy AI agents that could actually speak to Indian users — in Hindi, Tamil, Telugu, Kannada, Bengali, Marathi and 16 more languages — had no good options. Existing platforms were either English-first with bolted-on translation, or too technical for non-developer teams to operate.
The opportunity was clear: build a no-code AI agent platform that genuinely understood Indic languages, could deploy across every major channel, and could be powered by any leading AI model.
What I did
The visual workflow builder — the core of Lytespeed. A node-based canvas where teams can design, configure, and deploy AI agent workflows without writing code. Pre-built nodes for common tasks, custom workflow support for complex automations, and a drag-and-drop interface built for speed.
22 Indic language voice support — not translation on top of English, but native voice conversation support across India's major languages. WhatsApp, Email, Web, iOS, Android, Slack — the agents deploy across every channel users are already on.
Multi-model AI integration — the platform connects to OpenAI, Gemini, DeepSeek, and Llama. Teams pick the model that fits their use case, cost, and data requirements. No vendor lock-in.
A complete SaaS product — Free Forever tier for getting started, Pay As You Go for growing teams, Pro Plan for enterprises. Built, priced, and launched in four months.
Result
Lytespeed went from a blank canvas to a live, publicly accessible SaaS product in four months — with real pricing tiers, real integrations, and real language support that competitors with larger teams and longer runways hadn't cracked.
Stack & Tools
