Featured Project

EchoRecall: Voice-First Memory Agent

Voice-first assistive memory agent for the visually impaired — speech-to-speech interaction for object-location retrieval.

FastAPISpeech-to-TextOpenAI TTSSSMLMySQLPython

View on GitHub

Detailed Overview

An assistive AI agent designed for visually impaired users. EchoRecall allows users to verbally tell the system where they placed objects, and later ask where those objects are. The system uses Speech-to-Text to understand commands, stores memories in MySQL with vector indexing, and responds via OpenAI TTS with SSML for natural, expressive speech.

Problem Statement

Visually impaired individuals often struggle to remember where they placed everyday objects. Existing solutions require visual interfaces or physical labeling, which are inaccessible.

Technical Solution

Built a voice-first agent using FastAPI as the backend, Speech-to-Text for input, MySQL for persistent memory storage, and OpenAI TTS with SSML for natural voice responses. The system maintains a spatial memory of objects and locations, retrievable through natural speech.

Architecture & Data Flow

User speech → Speech-to-Text → FastAPI → Intent classification → MySQL memory store → Vector retrieval → OpenAI TTS with SSML → Audio response