Event
TIC Hackathon 2.0
🏆 TIC Hackathon 2.0 Winner
EchoMind
A full-stack voice-powered document Q&A system that lets healthcare professionals interact with medical documents through natural voice conversation. Built with RAG, OpenAI Whisper, and React in under 24 hours.
// overview
Project Overview
EchoMind is an award-winning voice-powered document Q&A system developed during TIC Hackathon 2.0 (October 24–25, 2025). The application combines speech recognition, vector embeddings, and large language models to create an intelligent assistant that can understand, search, and answer questions about uploaded documents.
Built with a focus on healthcare professionals, EchoMind lets users upload medical documents (PDF, DOCX, PPTX), ask questions using their voice, and receive spoken responses with source citations. The system uses Retrieval-Augmented Generation (RAG) to provide accurate, context-aware answers.
Duration
24 Hours
Team Size
4 Developers
Result
1st Place
// challenge
The Problem We Solved
Healthcare Information Access Challenges
Healthcare professionals often need quick access to information buried within large documents. Traditional methods require:
- Manually searching through lengthy PDFs and medical records
- Remembering which document contains specific information
- Typing out queries while multitasking with patients
- Reading through pages of text to find answers
Our Solution
EchoMind addresses these pain points by providing:
- Hands-free interaction: voice input/output allows querying while hands are occupied
- Semantic search: natural-language questions find relevant info automatically using RAG
- Contextual answers: the LLM synthesizes information into clear, conversational responses
- Source transparency: every answer includes citations showing exactly which documents were used
// capabilities
Key Features
Voice-First Interaction
Record voice questions using the browser microphone with real-time transcription via OpenAI Whisper. Natural text-to-speech responses with auto-silence detection.
RAG-Powered Search
Semantic search using OpenAI embeddings (1536 dimensions) with a ChromaDB vector database. Smart text chunking with overlap for context preservation.
Multi-Format Support
Upload PDFs, Word documents (.docx), and PowerPoint presentations (.pptx). Drag-and-drop interface supporting files up to 50MB.
Animated Voice Orb
Beautiful UI with 5 animated states: idle, session, listening, speaking, processing. Ripple effects and pulsing animations for visual feedback.
Secure Authentication
JWT-based auth with Individual and Organization account types. Role-based access control (owner, admin, member) with password-strength validation.
Three Query Modes
File Mode for document queries, Medical Knowledge for general questions, and Database Query for patient-records search.
// interface
Screenshots
Visual overview of the EchoMind platform interface.
// stack
Technologies Used
Backend
Python 3.11FastAPIPostgreSQLChromaDBSQLAlchemyLangChain
Frontend
React 19ViteTailwind CSSReact RouterAxios
AI & APIs
OpenAI Whisper (STT)OpenAI TTSOpenAI EmbeddingsTogether AI (Llama 3.1)
Infrastructure
VercelRenderJWT AuthFernet Encryption
// takeaways
What I Learned
1
Retrieval-Augmented Generation (RAG)
Implementing vector embeddings for semantic search, chunking strategies with overlap for context preservation, and balancing retrieval precision with LLM context windows.
2
Voice Interface Development
Integrating the browser MediaRecorder API for audio capture, working with OpenAI Whisper for accurate speech-to-text, implementing TTS with multiple voice options, and handling audio streaming in the browser.
3
Full-Stack Architecture Under Pressure
Designing REST APIs with FastAPI, managing state between the React frontend and Python backend, implementing JWT authentication with refresh tokens, and handling file uploads and processing pipelines — all in 24 hours.
4
LLM Integration Patterns
Prompt engineering for query refinement, two-pass systems for error correction, handling API rate limits and retries, and optimizing token usage and response quality.
5
Healthcare Domain Considerations
Understanding compliance considerations, the importance of disclaimers, and how to present medical information responsibly without overstepping into diagnosis.
// contribution
My Role
Backend Developer & Full-Stack Contributor
As the backend developer in our 4-person team, I was responsible for building the core infrastructure that powers EchoMind:
- Designed and implemented the RESTful API architecture using FastAPI
- Built the RAG pipeline with a ChromaDB vector database and OpenAI embeddings
- Integrated OpenAI Whisper for speech-to-text and TTS for voice responses
- Set up the PostgreSQL database schema with SQLAlchemy ORM
- Implemented secure JWT authentication and role-based access control
- Configured deployment on Render with managed PostgreSQL
- Collaborated with the frontend team on API contracts and data flow
Interested in Working Together?
I'm always excited to discuss new projects and opportunities in AI, healthcare tech, or full-stack development.