Inspiration
The idea for Aura came from a simple observation: people use multiple apps every day to manage different parts of their lives, calendars for planning, banking apps for finances, fitness trackers for health, journaling apps for mental wellness, and messaging tools for communication.
Yet, these systems don’t talk to each other.
I wanted to explore what would happen if a single intelligent system could connect all these aspects and act like a digital twin , an AI ecosystem that understands a person’s habits, routines, and preferences to provide meaningful assistance.
This project was also inspired by the growing trend of multi-agent AI systems, where different AI modules collaborate instead of relying on one monolithic model.
What it does
Aura is a multi-agent personal AI ecosystem designed to act as a digital twin for everyday life. Instead of using separate apps for planning, fitness, finances, and communication, Aura brings everything together through specialised AI agents that work collaboratively.
Each agent focuses on a specific life domain, such as scheduling, budgeting, nutrition tracking, workout guidance, mental wellness, and message summarisation — while sharing insights through a central memory and recommendation system. This allows Aura to provide context-aware, personalised support rather than isolated suggestions.
The goal is to create an AI that doesn’t just respond to commands but understands routines, adapts to habits, and coordinates different aspects of life intelligently.
How we built it
Aura was designed as a modular, multi-agent architecture.
The frontend is built using React Native, providing a mobile-first interface where users can interact with different agents through dashboards and conversational inputs.
The backend runs on Python with FastAPI, where each AI agent operates as an independent service. These agents communicate through shared data layers and a central memory system.
For the AI layer, we integrated:
Large Language Models for summarisation, conversation, and reasoning
NLP models for sentiment and text analysis
Computer vision models for food recognition in the Nutrition Agent
Used Supabase. This structure allows the system to be scalable, flexible, and ready for real-world integrations.
Challenges we ran into
One of the biggest challenges was designing coordination between multiple agents without making the system overly complex. Each module needed to remain independent while still sharing useful insights.
Another challenge was data consistency across domains. Financial data, fitness metrics, and text conversations are very different in structure, so creating a unified recommendation logic required careful abstraction.
We also had to think deeply about privacy and security, since the system concept involves highly personal user data. This influenced decisions around authentication, secure communication, and data handling practices.
Finally, managing the scope of such an ambitious idea was challenging. Turning a broad vision into a feasible, structured system required prioritising core features and designing for future expansion.
Accomplishments that we're proud of
We’re proud of successfully designing a true multi-agent AI architecture rather than a single monolithic assistant. Each agent has a defined role, making the system easier to scale and extend.
Another key achievement is creating a system that focuses on holistic personalization — combining signals from different life areas to generate smarter insights.
We also demonstrated strong skills across:
Full-stack mobile and backend development
AI model integration in real workflows
System architecture and modular design
Most importantly, Aura represents a step toward the future of personal AI ecosystems — intelligent systems that collaborate, learn, and grow alongside their users.
What we learned
Building Aura taught us that creating an intelligent system is not just about using powerful AI models — it’s about designing how different pieces work together.
We learned how to structure a multi-agent architecture, where each module has a clear responsibility but can still share information to create meaningful, cross-domain insights. This improved our understanding of scalability, modular design, and real-world AI integration.
We also gained hands-on experience in:
Connecting AI models to practical user workflows
Designing backend systems that support asynchronous and modular services
Handling diverse data types such as text, activity logs, and user preferences
Thinking critically about privacy, security, and ethical data use
Most importantly, we learned that personalization is not just about collecting data — it’s about interpreting context and turning signals into useful, human-centered recommendations.
What's next for AURA AI
Aura is designed as a foundation that can continuously evolve.
Next steps include integrating real-world APIs such as calendars, financial services, and wearable devices to move from a conceptual ecosystem to a fully connected personal AI platform.
We also plan to improve the Recommendation Engine, enabling deeper cross-agent intelligence so that insights from one area of life can meaningfully influence another — for example, adjusting schedules based on wellness data.
Future development will focus on:
Stronger personalization through long-term memory models
Smarter automation of daily routines
Enhanced privacy controls and transparent data handling
Expanding the voice and conversational interface for more natural interaction
Our long-term vision is to transform Aura into a true digital twin — an AI ecosystem that grows with the user, understands patterns over time, and provides proactive, trustworthy support in everyday life
Built With
- fastapi
- huggingface
- openai
- python
- react-native
- supabase