Personal Project · 2025
Zentric AI Health Companion
A holistic AI companion for chronic illness management, blending LangGraph orchestration with empathetic UX to transform healthcare support into a warm, human-centered experience.
Zentric Vision
Zentric is an AI-powered health companion designed to support people living with chronic illnesses through their daily care journey. Rather than functioning as a standalone chatbot, it combines health tracking, emotional wellbeing, personalized education, and conversational AI into one continuous experience.
The current MVP focuses on conversational intelligence powered by LangGraph, while the long-term vision expands into a complete mobile health platform with health dashboards, symptom tracking, glucose monitoring, wearable integrations, and personalized wellness insights.
The Challenge
Zentric exists because most health AI feels cold and clinical and for someone managing a chronic illness every day, that feeling makes an already hard situation harder. Most digital health applications focus heavily on data tracking while ignoring the emotional state of the patient, which makes manage Chronic illness more challenging.
Zentric was designed to solve the problem of accessible and non-judgmental support for people managing conditions like Type 2 Diabetes. The challenge was creating an AI system capable of handling complex workflows, biometric logging, recipe guidance, mindfulness support, and conversational memory while still maintaining a calm and empathetic tone.
The Strategy
I architected a state-driven agent workflow using LangGraph to manage non-linear conversations and persistent health context.
The system was decoupled into specialized nodes for data collection, RAG retrieval, conversational support, and tool-calling behavior. This allowed Zentric to intelligently route conversations depending on user intent while preserving the user's chronic condition context throughout the experience.
By combining Retrieval-Augmented Generation with tool-based orchestration, Zentric creates a more emotionally supportive and context-aware health companion.
The Zentric Platform
❤️ Daily Wellbeing
- Mood Check-ins
- Stress Reflection
- Energy Tracking
🩸 Health Tracking
- Glucose Logging
- Symptom Tracking
- Medication Reminders
🥗 Personalized Care
- Recipe Suggestions
- Meal Guidance
- Nutrition Coaching
🤖 AI Companion
- Health Education
- Daily Conversations
- Personalized Warm Guidance
Technical Execution
The backend architecture relies on a directed state graph built with LangGraph. A Fast API acts as the orchestration layer between the frontend interface, LangGraph workflow, RAG retrieval system, and external APIs.
The workflow begins at an initialize_profile_nodem which updates persistent user context and routes requests using a custom route_input conditional edge.
Specialized nodes were created for:
- Mindfulness-focused RAG retrieval
- Conversational emotional support
- Recipe generation through Spoonacular API tools
One of the largest technical challenges involved preventing infinite agent loops during tool execution. This was solved by introducing controlled state updates and stricter routing logic before returning control back to the conversational layer.
Beyond orchestration, significant effort was dedicated to tone engineering. The RAG knowledge base was intentionally curated to reduce overly clinical language and ensure responses remained emotionally warm, supportive, and psychologically safe for users navigating chronic illness.
Workflow Architecture
Daily Dashboard
│
┌─────────────┼──────────────┐
│ │ │
Mood Check Health Metrics Nutrition
│ │ │
└─────────────┼──────────────┘
│
AI Companion
│
Intent Recognition
│
LangGraph State Machine
│
RAG + External Tools
│
Personalized Guidance
│
Memory & User History
│
Updated Dashboard
Chat Agent Workflow
[User Input]
↓
[initialize_profile_node]
↓
(route_input)
├── collect_data_node → END
├── retrieve_rag_content → conversational_node → END
├── tool_agent_entry_node
│ ↓
│ call_model
│ ↓
│ call_tool
│ ↓
│ call_model ...
│
└── conversational_node → END
Gallery
Future Vision
The next evolution of Zentric focuses on persistent memory, wearable integrations, proactive health planning, and scalable RAG infrastructure that can be supported by real health specialists.
Future iterations may integrate wearable Fit applications integrations, medication tracking systems, nutritional estimation APIs, and persistent vector databases for long-term personalization.
The long-term vision is to evolve Zentric from a conversational assistant into a fully adaptive personal wellness agent capable of supporting emotional, behavioral, and physical health in real time.
Key Takeaway
High-performing AI systems are not enough in healthcare. Technology must also feel emotionally safe and human.
Zentric demonstrates how complex agent orchestration and state-machine architectures can be abstracted into a calm, supportive conversational experience that reduces stress instead of increasing it.
The Impact
Zentric represents a blueprint for next-generation personal health agents capable of combining emotional intelligence, contextual memory, and intelligent workflow orchestration.
The project explores how AI can evolve beyond transactional chatbots into persistent wellness companions that actively support long-term care journeys and help with a healthy lifestyle.