Omniscope · Data analytics · AI · Citizen science
I build software for making sense of data.
I’m Antonio Poggi, COO at Visokio and part of the team that builds Omniscope. I work across product, software and data, experiment with AI when it is genuinely useful, and apply the same tools to air-quality work around my hometown of Taranto.
COO at Visokio and part of the team behind Omniscope
What I work on
Most of my working life has gone into Visokio and Omniscope. The air-quality project began later as something personal: I wanted to make public monitoring data easier for people in Taranto to investigate and use. The subjects are different, but in both cases I care about making the data and the work behind it visible.
Building Omniscope
More than twenty years of product decisions, customer problems, prototypes, releases and engineering work around one complete analytics platform.
Air quality in Taranto
Public monitoring data, automated benzene alerts, work with local citizens and PeaceLink, and now an open sensor reporting privately into Omniscope.
Air quality · Taranto
I started with the public monitoring files.
The data was already being collected. I wanted citizens to be able to explore it, compare periods and see unusual events without first becoming data analysts. I later added automated benzene alerts, and PeaceLink used analysis from the project in civic work and at the 2024 Senate hearing.
Writing
16 articles based on things I have built, tested and learned.
What Is Verifiable AI Analytics?
Verifiable AI analytics separates model planning from governed execution, preserving the queries, calculations, lineage and evidence behind every answer.
02From CSV Prototype to Scheduled Data Application
A practical maturity path from a CSV analytics prototype to a parameterised, monitored, permissioned and scheduled internal data application.
03How to Build a Private AirGradient-to-Analytics Pipeline
Build a private AirGradient analytics pipeline using custom ESP32-C3 firmware, local Android control and direct HTTPS ingestion into Omniscope.
Three recent builds
These are quite different projects, but they show the kind of work I enjoy: start with a real question, build the complete path and leave enough behind for somebody else to inspect it.
A data-quality app in four hours.
A user uploads a CSV or Excel file and gets an interactive report on missing values, duplicates and outliers. ChatGPT helped with bounded pieces; I kept the architecture, assembly and deployment in hand. Coffee break included.
One question across five tables.
The AI produced its answer in 30 seconds. Inspecting the joins and transformations took another three and a half minutes, which I counted as part of the task.
An open sensor calling Omniscope directly.
I bought the hardware, reflashed the firmware, fixed the TLS failure, connected it to the workflow and built the Android controller over one weekend.