Data Scientist . AI & ML Engineer . Research Data Analyst
Amin AI Assistant - Session-Aware Conversational Intelligence
A modern AI assistant built to combine voice interaction, document ingestion, semantic memory, and fast API-driven chat into a single
practical workspace for research, support, and knowledge discovery.
Amin AI Assistant is a full-stack conversational assistant prototype that preserves session context, ingests uploaded documents, captures voice prompts,
and serves intelligent responses through a modern React chat interface backed by FastAPI.
Session-based chat with persistent conversation history and delete/restore controls
File upload ingestion for document-aware question answering
Browser-based voice recording with transcription routed into the chat flow
Streaming assistant responses for a smoother conversational experience
Core Capabilities
Context preservation across sessions for multi-step conversations
Voice-driven prompts converted to text and processed in the same chat history
Upload ingestion that transforms documents into semantic memory context
Hybrid UI with a lightweight React frontend and extensible Python backend
Quick session switching, active session highlighting, and conversation deletion
User Inputs
The interface supports multiple input modalities, making it simple to interact with AI using text, voice, or files.
Text entry for direct chat prompts and conversational queries
Voice recording with browser microphone capture and backend transcription
File upload for PDFs, text files, and other documents used to enrich response context
Clear status indicators for uploads, recordings, and session state
Modern chat interface with file upload and voice recording controls.
Assistant Output
AI-generated responses streamed to the UI for immediate feedback
Recorded audio playback after transcription with visible in-conversation status
Session-level summaries and context-aware follow-up responses
Chat transcript persistence for revisiting earlier conversation threads
Architecture & API
Backend: FastAPI provides `/chat`, `/chat-stream`, `/upload`, `/voice`, and `/sessions` endpoints
Frontend: Vite + React powers a responsive, session-driven chat workspace
Memory: local semantic memory stores paired with vector search for document retrieval
Voice: browser media capture uploads audio files to the backend for Whisper transcription
Session persistence: conversations are stored and selectable via sidebar session cards
Deployment & Security
Environment-driven configuration for API keys and runtime settings
Local gitignore rules for secrets, virtual environments, and generated artifacts
Designed for deployment on Render, Azure App Service, or other container-friendly hosts
Separate frontend and backend stacks to support independent build and release cycles
Business Impact
Reduces friction for AI interaction by combining text, voice, and file workflows
Improves productivity for research, support, and knowledge worker scenarios
Supports richer answers by incorporating document context into the conversation
Provides a strong portfolio example of full-stack AI product engineering
Tech Stack
Python, FastAPI, Uvicorn
React, Vite, TypeScript, modern frontend state management
Groq LLM API integration for core language generation
sentence-transformers + FAISS for semantic embedding and retrieval
Whisper for speech transcription and pyttsx3 / edge-tts for audio synthesis
