HackathonFeatured ProjectStatus: Award Winner38 Stars

Sahayog

Voice-first AI assistant concept designed to break digital literacy barriers for regional users.

My RoleHackathon Lead & Conversational AI Developer

Timeline / Duration36 Hours (Hackathon Sprint) + 2 Weeks Polish

Team Structure4 Members

THE CORE PROBLEM

Millions of elderly and regional language speakers in rural areas find complex form-based smartphone interfaces intimidating and inaccessible.

THE ENGINEERED SOLUTION

Sahayog provides a zero-touch, conversational voice interface that converts speech into structured actions (such as checking crop prices, setting reminders, or filing service queries) in conversational Hindi and Marathi.

PRODUCT CAPABILITIES

Key Features & Engineering Highlights

Continuous speech-to-text with auto noise cancellation for ambient outdoor use

Context-aware intent classification translating conversational speech into API payloads

Audio feedback synthesizing natural human-like responses in regional languages

Offline capability for critical emergency contact speed-dial commands

Technology Stack

ReactWeb Speech APIPythonFastAPIGemini AI APITailwind CSS

SYSTEM DESIGN

Technical Architecture

Microphone input is captured on the browser via Web Speech API, sent to a FastAPI inference engine that extracts intent with an LLM, and dispatches downstream microservices.

Voice Input / Output

Real-time audio visualization waveform and instant transcript feedback.

Web Speech RecognitionWeb Audio APIReact UI

Conversational Brain

Extracts user intent and required parameters (e.g. city, commodity, action).

FastAPIGemini AI APILangChain Intent Parser

Service Connectors

Fetches real-time commodity rates and public advisories.

Government Open Data APIsWeather APIsMarket Mandi APIs

DEEP DIVE CHALLENGES