-
-
Voice Agent
-
Voice Agent user flow
-
Baggage + Solana user flow
-
Tech/API work flow
-
Search places nearby
-
Baggage and Item Tracking Using proposed CV Pipelines in simulated airport environemnt
-
Claim or Report Missing Baggage Items
-
Live Tracking Of Baggage
-
Live Location Map Window
-
Inbound and Outbound Flights
-
Baggage item info sheet
#1 Best Overall Project -- CORTEX
the complete travel ecosystem.
Inspiration
I'm kind of blind. Standing at baggage claim, squinting at every black suitcase, hoping one's mine. I've lost more power banks and water bottles in airports than I can count. Set it down, walk away, gone.
And airline voice assistants? Garbage. Slow, robotic, painful.
We built CORTEX because we were tired of losing things, and tired of boring unhelpful agents.
Our 4 Features!
Item Tracking
Set down your water bottle or power bank, and we save the exact spot. Tap "find" and we guide you back. Like AirTag, without the tag.
Bag Detection
Scan your bag at check-in. When it hits the carousel, we ping you. No more standing around aimlessly.
Solana Integration
Every checked bag is tracked on-chain, giving flyers and airlines like American Airlines a transparent, tamper-proof record of where luggage is at all times. Decentralized nodes keep it working even with spotty connections while flying in the air.
Airlines lose over $2 billion annually to lost luggage and fraudulent claims. With CORTEX, users pay a small fee to check their bag and receive a refund plus a bonus upon retrieval, incentivizing honest travel and a possible new airline membership tier while protecting airlines from scams.
Voice Agent
Integrated with various airline APIs, ask anything about your flight. "Which gate do I need to go to? Where's gate B12?" Instant, human-sounding response. No hold music. No robot.
How We Built It
- iOS: Swift, ARKit, Vision, CoreML for hand detection and object tracking
- Web: Solana, Next.js, React, Three.js, Gaussian splatting for 3D terminal view
- Backend: Node.js, OpenCV for bag detection, WebSockets for real-time alerts
- AI/Voice: Gemini 2.0 Flash + ElevenLabs (STT and TTS)
- Infrastructure: MongoDB Atlas, Auth0, vultr
Challenges We Ran Into
Gaussian Splatting was hard the size of the auditrium, it took a lot of compute, which we ended up using vultr's GPU-accelerated virtual machines to help process our scans
Accomplishments We're Proud Of
- A CV pipeline that knows when you put something down
- Sub-2-second voice responses that sound human
- Bag detection that pings your phone automatically
- 3D Gaussian splat scan showing where you left your stuff
- 4 people. 24 hours. 4 features. No sleep.
What We Learned
- Gaussian Splatting is much more computationally heavy than I thought
- building a Gssian Splat renderer lwky hard (thank u vultr)
- ElevenLabs handles both STT and TTS has some interesting voices
Gemini 2.0 Flash is fast enough for real-time voice
Red Bull is goated
What's Next for CORTEX
- Real airline terminal integration
- More trackable items
- Proactive alerts before you board
- Apple Watch support
Built With
Gemini 2.0, ElevenLabs, Auth0, MongoDB Atlas, Solana, Vultr, Gaussian Splatting, OpenCV, ARKit, CoreML, Three.js, Swift, Next.js, React, Node.js, Tailwind, various airline APIs
Team
Ryan Ni, Lloyd, Ayad, Sri
CORTEX, We remember. You don't have to.


Log in or sign up for Devpost to join the conversation.