Project Submission: AuraSync (A.U.R.A. S.y.n.c.)

Inspiration

The inspiration for AuraSync came from a frustratingly simple real-world problem. The other day, a bright green line suddenly flashed across my monitor. It wasn't an issue I could solve through online shopping; I had to physically take it to an authorized service center in the electronics district. Unfortunately, I don’t live anywhere near that tech hub.

Every single time I actually remembered to deal with it, I was completely in the wrong place—either buried under project deadlines at my desk, or miles away from the tech hub. By the time I finally left the house and drove right by the repair market later that week, my mind was so fried from decision fatigue that I completely forgot the monitor was sitting in my back seat.

This made me realize the loop of modern information overload. Our brains are drowning in daily tasks, and traditional, static productivity apps are completely useless here. They expect you to stare at a rigid checklist and magically remember to do things, regardless of your active physical location or mental capacity. Online shopping cannot fix errands that require physical presence and real-time proximity. We didn’t need another boring to-do list; we needed an intelligent system that adapts to an individual's exact physical context, live locations, and focus boundaries. That is why we built AuraSync.

What it does

AuraSync (Automated User Real-time Assistant & Spatial Contextualizer) is a privacy-first, context-aware productivity engine that intelligently routes real-world tasks around your life. Instead of forcing you to check a static list, AuraSync operates silently in the background.

1. Contextual Text Parsing: When you input a task, the system maps out its underlying spatial and temporal intent without sending your personal data to a remote cloud server.
2. Focus-Boundary Safeguards: It respects your cognitive boundaries. If you are sleeping, working, or deep in a coding session, all intrusive notifications are completely muted.
3. Smart Push Notifications: The app calculates transitions into verified free time. It only fires an intelligent push notification when your calendar opens up and you are in real-time proximity to where that task can actually be completed.
4. The Rewarding Marketing System: It disrupts traditional, disruptive advertising by mathematically matching partner tasks exclusively with active user intent. When users complete highly relevant tasks—such as visiting a local partner tech store or service center—secure client-side GPS location verification instantly rewards them with Aura Points. These points convert directly into store coupons and gift cards, helping small businesses capture high-conversion foot traffic while maximizing the user's everyday earning potential.

How we built it

AuraSync was engineered using a local-first stack optimized for zero cloud latency and complete data sovereignty.

The task classification backend utilizes an embedded vector system to convert natural text descriptions into mathematical coordinates. For a given task description text $T$, we generate a high-dimensional dense vector representation using local text embeddings:

$$\vec{v} = \text{Embed}(T)$$

The application then evaluates similarity against pre-mapped physical contexts or vendor intents represented by coordinate vectors $\vec{c}$ within our local multi-dimensional index. This mapping determines context by evaluating the cosine similarity metric:

$$\text{Similarity}(\vec{v}, \vec{c}) = \frac{\vec{v} \cdot \vec{c}}{|\vec{v}| |\vec{c}|} = \frac{\sum_{i=1}^{n} v_i c_i}{\sqrt{\sum_{i=1}^{n} v_i^2} \sqrt{\sum_{i=1}^{n} c_i^2}}$$

If the similarity score exceeds a defined threshold ($\text{Similarity} > \tau$), the task is automatically assigned to that physical boundary vector.

On the frontend, we built the interface utilizing cross-platform components, integrating native modal animations, and a secure client-side GPS listener. This background script continuously checks real-world geographic coordinates against target zones without exposing or transmitting raw coordinate histories to third-party ad networks.

Challenges we ran into

Building an intent-matching machine that works strictly client-side brought complex engineering bottlenecks:

• Local Embedding Constraints: Running text tokenization and vector mathematical indexing directly on mobile and low-power hardware caused initial lag. We resolved this by aggressively optimizing our indexing math and utilizing compressed, quantized embedding models.
• Battery-Efficient Proximity Tracking: Continuous GPS polling drains mobile batteries rapidly. To prevent this, we built a nested tracking system: we used low-energy cell-tower geofencing to detect major regional shifts, and only spinning up the high-precision hardware GPS chip when the user entered a bounding box close to a target partner zone.
• Notification Spam Fatigue: Initially, entering a market area would trigger multiple spatial notifications at once. We had to write an explicit cooling-off algorithm based on user availability to ensure that notifications are strictly serialized and pushed only when calendar availability changes to "Free Time."

Accomplishments that we're proud of

• Zero Cloud Latency: The entire vector-parsing and matching pipeline runs locally. Task processing takes milliseconds, keeping user data entirely within their own device boundaries.
• High-Intent Marketing Engine: We successfully proved that advertising doesn't have to be invasive. By mapping partner stores to explicit user needs, the marketing system behaves like an organic recommendation rather than a pop-up ad.
• Smooth UX Interface: We built a gorgeous, fluid UI with clean modals that present task metrics, location verifications, and user rewards seamlessly without overwhelming the user.

What we learned

• Local-First is Viable: We learned that with modern optimization techniques, mobile devices are fully capable of handling dense vector math and text parsing locally, eliminating the need for expensive cloud hosting.
• Context is King: Productivity isn't about working harder; it's about minimizing the friction of remembering. Aligning cognitive capacity, time, and physical space yields far better task-completion rates than rigid time-blocking.
• Privacy Controls Incentivize Engagement: Users are much more willing to participate in proximity rewards and location check-ins when they know the data never leaves their local device sandbox.

What's next for Automated User Real-time Assistant & Spatial Contextualizer

The journey for AuraSync is just beginning. Moving forward, we plan to implement:

• Multi-Modal Parsing: Allowing users to snap photos of broken items (like that green line on my monitor) or physical flyers, running local computer-vision models to automatically extract task items.
• Decentralized Partner Ledger: Moving the Aura Points validation ledger onto a lightweight, cryptographic network to allow trustless, instantaneous reward coupon verification between global corporate brands and independent local businesses.
• Predictive Routine Modeling: Enhancing our temporal engine to map out historical patterns safely, predicting when a user is likely running casual errands versus working, making notification schedules even more accurate.

Built With

• css
• html
• javascript
• react
• supabase
• typescript
• vite