The AI stack behind every scan
Tincyglo isn't one model — it's a pipeline of computer vision, deep learning, and generative AI, running on infrastructure built to stay accurate at scale.
68-pt
Facial landmark mesh
8
Independently scored skin markers
97.4%
Marker detection accuracy
<1s
Median scan processing time
Every scan runs through four AI disciplines
Computer Vision
Extracts texture, tone, and surface detail from a single 2D image at pixel-level resolution.
Machine Learning
Classical ML models score discrete markers like redness and oiliness from engineered skin features.
Deep Learning
Convolutional networks trained on dermatologist-labeled imagery detect acne, wrinkles, and pigmentation patterns.
Generative AI
A generative layer turns raw scores into routines, explanations, and conversational coaching.
Face Landmark Detection
68-point facial mapping normalizes every scan for angle, distance, and lighting before analysis.
Image Processing
Pre-processing pipelines correct exposure and color balance so results stay consistent across devices.
Medical AI
Models are validated against dermatologist-reviewed datasets, not just consumer photo sets.
LLMs
Large language models power the AI Skin Coach, grounded in each user's own scan history.
Built to stay accurate, secure, and fast at scale
Cloud Infrastructure
A fully managed, auto-scaling pipeline processes scans in under a second, globally.
AWS
Built on AWS — S3, SageMaker, and Lambda — for elastic compute and durable storage.
Security
Encryption in transit and at rest, with role-based access across every environment.
Scalability
The same pipeline serves a single consumer scan and a hospital's bulk-upload batch job.
Accuracy
Every model release is benchmarked against a held-out validation set before shipping.
Want the technical deep dive?
Our AI Research page covers datasets, validation methodology, and responsible AI practices in full detail.