Data Scientist . AI & ML Engineer . Research Data Analyst
Sentinel AI — Intelligent Multi-Agent WhatsApp Assistant
Sentinel AI is a production-oriented conversational AI platform that
combines Large Language Models, LangGraph, FastAPI, Supabase PostgreSQL,
document intelligence, speech recognition, computer vision, persistent
memory, and web search into a single intelligent assistant accessible
directly through WhatsApp.
Sentinel AI is a modular AI assistant engineered to demonstrate
production-grade AI engineering principles rather than a simple
chatbot implementation. It integrates conversational intelligence,
persistent memory, external tools, document processing, and cloud
infrastructure into a scalable backend architecture suitable for
intelligent automation and business applications.
Production-ready FastAPI backend
LangGraph multi-agent orchestration
Persistent conversational memory
WhatsApp Cloud API integration
Modular architecture for future expansion
Core Capabilities
Natural language conversations powered by Groq Llama 3.3 and Google Gemini
Persistent user memory backed by PostgreSQL
Real-time internet search for up-to-date responses
PDF document understanding and summarization
Spreadsheet profiling and AI-driven insights
Image understanding using Gemini Vision
Voice message transcription and conversational response
User authorization and duplicate webhook protection
Structured logging and production-ready backend architecture
AI Workflow
Every incoming WhatsApp request flows through multiple intelligent
processing stages before a response is generated. This architecture
enables context-aware conversations while maintaining scalability,
security, and extensibility.
WhatsApp Cloud API receives incoming user messages
FastAPI webhook validates and processes requests
Authorized users are verified before execution
Duplicate webhook events are ignored automatically
LangGraph routes requests to the appropriate AI workflow
Memory retrieval enriches conversations with previous context
External tools are invoked only when necessary
LLMs generate context-aware responses
High-level workflow illustrating the complete request lifecycle.
Memory & PostgreSQL
Unlike conventional chatbots that forget previous interactions,
Sentinel AI maintains persistent conversational memory using
Supabase PostgreSQL. User-specific information is stored and
retrieved intelligently, allowing conversations to remain
personalized across multiple sessions.
Supabase PostgreSQL cloud database
Persistent user profiles
Long-term conversational memory
Duplicate message tracking
Conversation continuity across sessions
Scalable relational database architecture
Persistent memory and conversation storage using Supabase PostgreSQL.
Intelligent memory retrieval powering personalized conversations.
Document & Multimodal Intelligence
Sentinel AI extends beyond conversational AI by supporting
multimodal interactions. Users can upload PDFs, spreadsheets,
images, and voice messages directly through WhatsApp and receive
AI-generated insights.
PDF extraction and intelligent summarization
Question answering from uploaded documents
Spreadsheet profiling and statistical summaries
Automatic data quality observations
Image understanding using Gemini Vision
Voice transcription powered by Whisper
AI-generated recommendations and explanations
AI-powered document understanding and summarization.
Spreadsheet profiling and intelligent dataset insights.
System Architecture
Sentinel AI adopts a modular architecture where each component is
responsible for a single concern. This separation simplifies
maintenance while making future integrations significantly easier.
FastAPI REST backend
LangGraph agent orchestration
Supabase PostgreSQL persistence layer
SQLAlchemy ORM
Dedicated service modules
WhatsApp Cloud API integration
Structured logging with Loguru
Environment-driven configuration
High-level system architecture of Sentinel AI.
Deployment & Security
Environment-variable driven configuration
Supabase cloud database deployment
Railway deployment pipeline
Authorized user access control
Webhook duplicate protection
Persistent structured logging
Production-ready FastAPI backend
Cloud-ready modular architecture
Business Impact
Automates customer interactions directly from WhatsApp
Reduces response time using intelligent AI workflows
Provides personalized assistance through persistent memory
Transforms documents and spreadsheets into actionable insights
Demonstrates scalable enterprise AI architecture
Suitable foundation for CRM, internal assistants, and business automation
Technology Stack
Backend: Python, FastAPI
AI Framework: LangGraph, LangChain
Large Language Models: Groq Llama 3.3, Google Gemini
