Federico Rao
PROJECTS
Selected systems spanning learning platforms, e-commerce, AI product tooling, analytics, CRM, and conversion-focused web applications.
How to read these records
Each project page describes an implementation: the problem it addressed, the architecture chosen, the workflows it supports, and the constraints that shaped it. Visible product behaviour, interface decisions, and technical structure can be inspected directly. Business outcomes that depend on client data are not published as proof, and none of these records should be read as a claim about revenue, growth, or performance for the organisations involved. Ownership is stated per project, including collaborative work and internal tools built without a client.
ELMS Learn, enterprise learning platform
A learning management system built as fourteen independently deployable Spring Boot microservices behind a React frontend covering the educational lifecycle: enrolment, course delivery, assessment, mentoring, placement, and analytics. Services communicate over REST and Kafka event streaming, each owning its own PostgreSQL database to enforce bounded contexts. The frontend uses role-based routing across four user types with code splitting and a resilient try-API-then-mock development pattern. It is the largest distributed-systems record on the site and the clearest example of multi-service architecture rather than a single application.
Acquamarina, commerce storefront
A bilingual luxury storefront for handcrafted goods, built around product browsing, cart, checkout, and authentication with a Mediterranean design system. State is handled with feature-scoped stores, forms are validated with schema-based typing, and the cart drawer uses spring animation. Full English and Italian localisation is part of the build rather than a later addition. It demonstrates commerce interaction design and internationalisation on a small, self-contained codebase.
Atelier Engine, AI imagery workflow
A desktop-class tool for generating product photography variants for fashion inventory. A vision model analyses source photographs, prompts are constructed from that analysis, and generation runs in batches producing studio, close-up, and on-model variants. The backend streams progress over WebSocket; the interface provides prompt authoring, run management, and workflow control. Operational concerns dominate the design: CSV inventory batches, pausing and resuming runs, recovery after interruption, and append-only audit logging.
MTRM, fashion trend analytics
A machine-learning pipeline that ingests social media text and images to score emerging demand signals in fashion. A custom text classifier trained on a demand lexicon handles language, an object-detection model identifies clothing items in screenshots, and scored output is served through an API over a document store. It is a research-oriented system: the interesting parts are the labelling approach, the scoring method, and the deployment shape rather than a polished consumer interface.
Praxis, practice management
A focused management interface for professional clinics covering the full case lifecycle: creation, active handling, notes and review, archival, and reactivation. The interface is deliberately narrow, with two panels separating active from historical cases, file attachments, and inline notes. It was built as a minimum viable tool for immediate deployment, replacing spreadsheets and paper files for one practice, and it illustrates how small a system can be while still being useful.
Third Eye, mobile audio product
A cross-platform mobile application for mindfulness and ambient sound mixing, layering ambient layers over master tracks through a custom interface. The client uses persisted state and GPU-accelerated graphics for mood-reactive visuals; the backend handles authentication, catalogue, mixer state, subscriptions, and content administration. It is the most complete mobile record here, covering onboarding, listening journeys, subscription handling, and the administrative surface needed to keep content current.
Page focus
Browse implementation records by product surface, workflow, technical constraint, and inspectable artifact rather than unsupported outcomes.
Portfolio method
Choose records by implementation problem and inspectable artifact; descriptions do not substitute for verified customer results.
Measurement and limits
Evaluate Web, automation, and AI projects with observable checks relevant to this route. Record material changes when the underlying offer, system, evidence, or dependency changes. Technical eligibility or deployment alone does not guarantee rankings, traffic, enquiries, revenue, accessibility compliance, or operational improvement.