Portfolio
.png)
Type : Personal proposal with a working detection prototype.
On trails, climbs, or remote trips, you or a bystander may be the only aid. Most untrained people freeze in trauma.
The kit uses AR and Machine Learning for instant diagnostics and spatial guidance, optimizing the critical "Platinum Minutes" for untrained bystanders. It monitors, guides and alerts the EMS about an injurious incident.
Aurum is an AR first-aid assistant that detects wounds and kit items (blood, laceration, hands, tourniquet, gauze, syringe) in real time using a YOLOv8 model trained on ~10,000 hand-labeled Pascal VOC images.
Detections drive a severity-tiered heads-up interface, with a TensorFlow Lite inference pipeline scoped for Meta Ray-Ban glasses and a Figma-prototyped UI for Meta Quest 3.
Role : Researcher, UX/UI designer, dataset and model builder.
AURUM
Design Target: Wearable AR glasses,
e.g - Meta Ray-Ban Display
When help is hours away, severe bleeding can be fatal.....
Can glasses recognize the injury quickly enough to choose the right guidance?
On remote trails, climbs and expeditions, the person next to the injured hiker is often the only responder, and usually has no training. Even where rescue helicopters exist, the transport clock is long.
In a Swiss alpine study of time-critical patients, these were the times from alarm to hospital:
-
A Bleeding uncontrollably can die in about 5 minutes.
Major concerns :
-
Remote means slow EMS response, which can cause further delay in medical treatment.
-
Denali and NASAR guidance both warn that wilderness phone coverage is variable, so the help can be delayed.
Problem statement: There is no system that supports untrained bystanders in real time, under stress, when help is far away.
33% under 60 mins
35% up to 90 mins
32% over 90 mins
Fear of causing harm
Bystanders hesitate for reasons an interface can address
Fear of causing harm
Persona Mapping
Injured traveler
Conscious, limited mobility
Age: 18-60
Goal : Survive, Get help
Pain : Isolated, In-pain, panic , No medical experience
Fear of causing harm
Basic First-Aid, Responsible
Travel Guide
Age: 25-50
Goal : Manage incident quickly
Pain : Liability, Coordination
Bystander/companion
Panicked, No medical training
Age: 18-60
Goal : Get help quickly
Pain : Panic, no medical experience
Sources - https://pmc.ncbi.nlm.nih.gov/articles/PMC12315186/
- https://www.who.int/news-room/fact-sheets/detail/injuries-and-violence
- https://www.who.int/news-room/fact-sheets/detail/injuries-and-violence
https://ec.europa.eu/health/ph_projects/2001/injury/fp_injury_2001_frep_10_en.pdf
Injury Stats
~4.2 million
Target users
714000
608 million
Travel related injuries
22000
Adventure travel injuries
Global injury cases
Fatal travel injuries
Nature of Injuries during adventure travel
Sources - https://www.thelancet.com/journals/lanhl/article/PIIS2666-7568(21)00172-0/fulltext
- https://www.sciencedirect.com/science/article/abs/pii/S003335062400252X
- https://academic.oup.com/jtm/article-abstract/31/1/taad088/7219761
What it sees : Injury , Kit , Context , Blood, Cut , Laceration , Syringe, tourniquet, gauze , Hands
What it does : Detects the nature and severity of injury
Data: 10,000 images labelled , trained , validated and tested
Data training : Collection of images > Clean the data ? Labelling XML files > Exports tflite file
Over 10,000 images collected for training, testing and validating
Training class syringe, to create XML file,
Collected datasets from Roboflow
Detection accuracy :
Detection classes with >50% accuracy
Fails to detect certain types of torniquets
Detection failure :
The interface is curated for one instruction at a time
UI flow for Abrasion
For a surface wound as such, EMS is not required yet. Therefore the user is instructed to clean, rinse and monitor the injury for further assistance.
UI flow for Bleeding Laceration
For a bleeding injury, it is important to clean, sterilize and treat it at its source, as uncontrolled blood loss can have severe consequences. Therefore the user is instructed to take best possible measures while EMS arrives. The user is guided to remain calm and composed during this.
Design Choices
Meta RayBans
Weight: 50g 

No Display.Audio Only.

Meta wearable SDK gives limited access or compatibility

The hardware gap. No single device in this comparison fits a wilderness emergency. Headsets give full passthrough but are bulky, lack GPS, or have short battery life. Optical see-through headsets have narrow fields of view. Lightweight glasses are wearable but offer only a small display and depend on a phone. Aurum's interface is built to work within that constraint.
Hardware gap
Meta Quest 2
Weight: 500g
No colour pass-through.

Needs tethering.

Limit battery life
Meta Quest 3
Weight: 500g 

Colour pass-through.

Limit battery life.

Limited developing options