BeautyAI Services

AI-Powered Skin Analysis Platform

Vision across 20+ skin parameters, recommendations and a beauty copilot for a Singapore skincare startup.

Key Details

Challenge	A Singapore skincare startup needed skin analysis, product recommendations and a conversational copilot in one product.
Solution	A computer-vision engine for 20+ skin parameters, an AI recommendation engine, an LLM beauty copilot, a progress dashboard and REST APIs.
Technologies	Vision Transformer, YOLOv8, OpenCV, PyTorch, ONNX Runtime, Azure OpenAI GPT-4o, LangChain, RAG, FastAPI, NestJS, React, React Native, AWS, Docker, Kubernetes

Technologies used

Vision TransformerYOLOv8OpenCVPyTorchONNX RuntimeAzure OpenAI GPT-4oLangChainRAGFastAPINestJSReactReact NativeAWSDockerKubernetes

Client

The client is a startup skincare and beauty brand in Singapore. Shoppers photograph their skin on a phone and receive an analysis, a routine and a beauty copilot.

Product screenshots

Two peach-toned phones: the first shows a selfie with a face mesh outline and highlighted analysis zones on the forehead, cheeks, nose and chin with small markers over detected spots; the second shows the results screen with an overall skin score ring and per-concern bars for spots, redness, texture, oiliness and hydration. — Skin scan result: the selfie is mapped into facial zones, and each zone gets a score for concerns such as spots, redness and texture.

Lilac web app with a chat on the left where the user asks why their cheeks show spots and the AI assistant answers with reference to their scan zones and suggests ingredients, and a routine builder on the right listing morning and evening steps with product types, a weekly calendar and a reminder toggle. — Skin copilot: the user asks follow-up questions about their scan and the assistant explains the result and builds a morning and evening routine from it.

What we built

Computer-vision engine for 20+ skin parameters, including acne, wrinkles, pores, pigmentation, redness, dark circles, texture, oiliness, hydration indicators and skin tone
Recommendation engine for skincare routines and products from the detected conditions
LLM beauty copilot that explains results and answers skincare questions
Customer dashboard that tracks skin progress from historical analysis
REST APIs for mobile apps and e-commerce
Analysis from a selfie taken on a smartphone camera

Engagement

Delivery was time and materials with 7 engineers over 12 months.

Project team: 7 engineers. Duration: 12 months. Model: Time & materials.

Services used

AI Services →DevOps →AI-Software Development →Document & Knowledge AI →