Project Details
Neluma
Neluma is an AI-powered beauty technology app concept that uses image analysis to support skin tone, undertone, and shade-matching workflows for users and makeup artists.
Overview
What the project does
Neluma focuses on helping users better understand their skin tone and undertone through image-based analysis. The project explores how computer vision, backend APIs, and structured user profiles can support more personalized beauty recommendations.
The app concept is designed around two main user groups: clients who want shade guidance and makeup artists who need a faster way to view, save, and manage client analysis results. The current work focuses on the technical foundation: image upload, backend analysis, shade matching, profile planning, and client/MUA workflow design.
To protect the full product idea, this case study focuses on the technical implementation and general workflow rather than every planned feature or business detail.
Product Concept
How the app is structured
Neluma is planned as a mobile-first app with separate experiences for regular users and makeup artists. A client can complete a skin analysis, view results, save analysis history, and use their profile for future shade guidance. A makeup artist can analyze clients, view recent client profiles, and manage beauty-related workflows from a more professional dashboard.
The UI planning includes screens for skin profiles, starting or retaking an analysis, viewing results, recommended products, history, profile settings, and MUA-focused client management. The purpose is to make the analysis feel simple for users while keeping the backend data useful for future recommendations and professional workflows.
Technical Flow
How the analysis workflow works
Image Upload
The user uploads or captures an image. The image is sent to the backend through an API request for analysis.
Backend Preprocessing
The Python backend preprocesses the image using tools such as OpenCV, NumPy, and Pillow to prepare it for analysis.
Face Detection
MediaPipe concepts are used to support face detection and identify relevant facial regions for skin tone analysis.
Skin Region Extraction
The system focuses on selected skin regions instead of analyzing the entire image. This helps avoid background colors, hair, clothing, and lighting noise.
Skin Tone + Undertone
The backend calculates skin tone values, determines undertone categories, and prepares structured analysis results.
Shade Matching Output
The analysis output can be used to support shade recommendations, profile updates, and future beauty-related workflows in the app.
Backend Work
What I built technically
I built and tested a FastAPI backend for the analysis flow. The backend accepts image uploads, runs preprocessing logic, detects whether a face is present, extracts skin-related regions, calculates skin tone information, classifies undertone, and returns shade-matching results in a structured response.
I organized the backend into services and utility files so the project could be easier to expand. This included separating image preprocessing, face detection, skin analysis, shade matching, and database-related logic instead of keeping everything in one file.
I also tested the API through local requests and connected early frontend test screens to make sure the iOS app could receive analysis results from the backend.
Database Planning
How user and analysis data would be stored
I explored Supabase as the database and storage layer for the project. The planned database structure includes user accounts, skin profiles, analysis history, uploaded image references, saved product matches, and permission-based sharing between a client and a makeup artist.
The goal is not only to run one analysis, but to support a reusable profile system where users can save results, compare updated analyses, and share selected information when needed. For a production version, privacy controls and user permissions would be important because the app handles personal image and profile data.
Client + MUA Flow
How the two-user workflow is planned
Neluma is planned with both client and makeup artist use cases in mind. A client can complete an analysis and view their personal results, while a makeup artist can use client-approved information to prepare for appointments, view past analyses, and manage beauty workflows more efficiently.
The MUA-focused planning includes client lists, recent analyses, shared profiles, portfolio planning, team-related ideas, and upgrade/payment flow exploration. These details are still in planning, so the public case study only shows the general workflow without revealing the full product strategy.
My Contributions
What I worked on
Backend API
Built FastAPI endpoints to process image uploads, run analysis logic, and return structured results to the frontend.
Image Processing
Used Python, OpenCV, NumPy, Pillow, and MediaPipe concepts to support preprocessing, face detection, and skin-region extraction.
Shade Matching Logic
Worked on undertone classification and foundation shade recommendation logic using structured analysis outputs.
API Testing
Tested backend routes, image uploads, response formats, and different image examples to check that the system returned usable results.
iOS App Planning
Planned SwiftUI screens for onboarding, analysis results, user profiles, history, product recommendations, and MUA/client workflows.
Database Architecture
Planned Supabase tables, storage needs, analysis history, profile data, and permission-based sharing for future app features.
Tech Stack
Tools used
Status
Backend prototype and app planning in progress.
Neluma is currently being developed as an AI-integrated beauty technology app with image analysis, shade matching, user profile planning, and future MUA/client workflows. The current work focuses on building a reliable technical foundation while keeping the full product concept private.