“When complexity isn't in the data. It's in the relationships around it.”
The challenge
In finance, data isn't simply information: it's a responsibility. The challenge isn't handling a single piece of data, but ensuring consistency and reliability in an ecosystem that grows over time — thousands of profiles and activities interwoven across four years of project.
Xeedom integrates profiling, scoring, external sources and decision processes. The challenge wasn't building the system, but designing an architecture able to preserve integrity, security and reliability over time.
The solution
Not a list of functions: three architectural choices that move security and correctness from the “final check” to the heart of the system.
Data that never even leaves
Authorization logic lives entirely on the server: unauthorized data is never sent to the client. No front-end filters to bypass — what isn't yours doesn't leave the backend.
Understanding relationships, not just fields
The rating doesn't come from a table: it rests on a Neo4j knowledge graph that models the relationships between entities, fed by a proprietary ML engine. Profiling is detailed and engaging, with a gamification layer.
The edge case as thesis
Integrations don't fail on the standard cases. They fail on the exceptions. The Infocamere integration returned, for a province change, “two companies” on an undocumented structure: we resolved it upstream, at the architectural level, before it could propagate. Above it, an engine that maps questions and fields, guides completion and crystallizes the data.
The architecture
Xeedom is an extended system: the front displays, but every decision on data, permissions and scores stays in the core. Eight components working together — from granular authorization to external integrations — without ever exposing what isn't due.
Profiling with gamification, data room and dashboard. It only receives data already authorized by the server.
Authorization evaluated before serialization, with granular access levels per role, entity and individual field. Unauthorized data never leaves the backend.
Document archive with customizable foldering, per-folder permissions and access tracking.
Models the relationships between subjects, companies and data: the map everything else reasons on.
A different goal from the graph: it learns from data to produce predictions and signals, not just to describe relationships.
A dedicated rating engine that combines graph and ML into a coherent, explainable score.
Question ↔ field mapping, guided tour, completion and crystallization of the collected data.
Normalized external source: edge cases (e.g. province change) resolved upstream, undocumented structure tamed.
The proof
We show screens and application flows, keeping sensitive data and information confidential. Technical deep-dive and due diligence available under NDA.
Screens from a demo environment (fictional data) · product link and DD available under NDA.
Let's talk