Guglielmo_Celata
← All work

A public-affairs consultancy · via DEPP · 2024 →

An LLM-based monitoring platform, in production

Analysts monitor the news stream for each client's interests; the volume makes manual triage the bottleneck. I designed an application that reads the stream and categorizes, summarizes and prioritizes it against each client's declared interests.

Fig. · ArchitectureIngest → reason → rank
News stream continuous Ingestion structured · governed LLM · via API › categorize › summarize › prioritize vs interests the layer designed for it Ranked feed per client → analysts act

My role

Ideation and application design, delivered within DEPP's offer. The first LLM-based application in the portfolio.

What it does

LLMs via API over an ingestion pipeline: it categorizes incoming items, summarizes them, and extracts and prioritizes what matters against each client's interests.

Why it matters

The production proof of the AI-application strand — and a concrete illustration of the nexus: the application works because the data layer beneath it (ingestion, structure, governance) was designed for it.

← All work