AI Skin Advisor & Product Recommender (FOREO Ecosystem)
AI Skin Advisor is an intelligent system based on large language models (LLM), designed to provide personalized skincare advice and automated product selection from the FOREO ecosystem. The project operates on the principle of a dialogue assistant that analyzes individual user queries, the condition of their skin, and offers optimal solutions for daily care.
Key functionality of the system
- Intelligent chat interface: Implementation of a chat window in the style of ChatGPT, allowing users to receive instant answers to questions about dermatological care.
- Personalized diagnostics: Analysis of user input data (skin type, age, current issues) to create a unique "digital profile" of the skin.
- Smart FOREO recommendations: Algorithmic selection of gadgets (LUNA, UFO, BEAR) and brand cosmetics, integrated into a personalized routine.
- Educational content: Generation of instructions and tips for the proper use of microcurrent technologies, T-Sonic pulsations, and LED therapy.
Technology stack and architecture
- NLP/LLM: Use of GPT-4 or Claude models for conducting contextual dialogues.
- Data Science & Machine Learning: Development of product ranking algorithms based on ingredients and technical specifications of devices.
- Backend: Python (FastAPI/Flask) for processing requests and integrating with the product database.
- Frontend: React/Next.js or integration into the FOREO For You mobile application.
- Data Storage: Vector databases (Pinecone/Milvus) for quick retrieval of relevant dermatological knowledge (RAG architecture).
My contribution (Data Science)
Within the project, I was responsible for the analytical part and the development of recommendation logic:
- Development of a recommendation model: Creating a mapping system between users' dermatological issues and the specifications of FOREO product lines.
- Fine-tuning and Prompt Engineering: Adjusting the behavior of the language model to ensure high accuracy of medical advice and adherence to the brand's tone of voice.
- Data processing and structuring: Formation and preparation of datasets with cosmetic ingredients and technical parameters of gadgets for training algorithms.
- Validation of responses: Development of metrics to assess the safety and relevance of the assistant's advice to avoid erroneous dermatological recommendations.
Key functionality of the system
- Intelligent chat interface: Implementation of a chat window in the style of ChatGPT, allowing users to receive instant answers to questions about dermatological care.
- Personalized diagnostics: Analysis of user input data (skin type, age, current issues) to create a unique "digital profile" of the skin.
- Smart FOREO recommendations: Algorithmic selection of gadgets (LUNA, UFO, BEAR) and brand cosmetics, integrated into a personalized routine.
- Educational content: Generation of instructions and tips for the proper use of microcurrent technologies, T-Sonic pulsations, and LED therapy.
Technology stack and architecture
- NLP/LLM: Use of GPT-4 or Claude models for conducting contextual dialogues.
- Data Science & Machine Learning: Development of product ranking algorithms based on ingredients and technical specifications of devices.
- Backend: Python (FastAPI/Flask) for processing requests and integrating with the product database.
- Frontend: React/Next.js or integration into the FOREO For You mobile application.
- Data Storage: Vector databases (Pinecone/Milvus) for quick retrieval of relevant dermatological knowledge (RAG architecture).
My contribution (Data Science)
Within the project, I was responsible for the analytical part and the development of recommendation logic:
- Development of a recommendation model: Creating a mapping system between users' dermatological issues and the specifications of FOREO product lines.
- Fine-tuning and Prompt Engineering: Adjusting the behavior of the language model to ensure high accuracy of medical advice and adherence to the brand's tone of voice.
- Data processing and structuring: Formation and preparation of datasets with cosmetic ingredients and technical parameters of gadgets for training algorithms.
- Validation of responses: Development of metrics to assess the safety and relevance of the assistant's advice to avoid erroneous dermatological recommendations.
Warszawa 
Available for hire