Evolve365
Prototype enterprise AI search platform: ingest mixed document sources into a vector knowledge base and answer grounded questions with an OpenAI-backed model.
What the prototype had to prove
Enterprise teams already hold knowledge across files, exports, and operational documents that never sit in one searchable system. Evolve365 was a Localhost prototype for evolving an enterprise AI search platform into a knowledge base fed by those mixed sources, so staff could ask grounded questions instead of hunting through folders. Public detail stays limited: the engagement was commercially sensitive.
What we built
A ChatGPT-style query surface on a Next.js frontend, backed by a FastAPI ingestion and retrieval service. Documents could be uploaded across common enterprise formats, embedded with OpenAI models, stored with vector search, and retrieved as context for answers. Identity used OIDC so access sat inside an organisation login path rather than a shared key.
What remains as delivery evidence
The useful residue is the shape of the work: source ingestion, retrieval quality, model-backed answers, and auth boundaries treated as one system. There is no public demo linked from this page.