IT Press Tour 69: Veridion

Veridion presented itself during the 69th edition of The IT Press Tour as a data company rather than a software vendor, with the core asset being a continuously updated knowledge graph describing companies worldwide.The platform currently covers 693 million companies across 250+ countries and territories, with 461+ attributes across 24 dimensions for each entity. These include locations, company size, business activity, corporate structure, M&A, technologies, products and services, risk signals, ESG, sustainability, supply chain and key people. The company argues that traditional firmographic databases remain heavily dependent on registries and manually maintained information, resulting in data that is often incomplete, stale and difficult to scale.

Click to enlarge

Founded as Soleadify in 2019 and renamed Veridion in 2023, the company employs more than 60 people, with engineering centered in Bucharest and commercial operations in North America and the UK. It has raised $7.5 million in two seed rounds and says revenue is currently growing at more than 2× YoY. Its customers include Experian, S&P Global, Moody’s, Morningstar, Dun & Bradstreet, FactSet, Hiscox and others. Veridion claims its approach delivers 30% higher accuracy, 50% more data points per company and roughly 50% lower cost than legacy providers.

Click to enlarge

The key differentiation is how the data is collected and maintained. Veridion crawls around 1.4 billion web pages per month, mostly using its own crawl infrastructure, and generates around 100 million model predictions per day. Its pipeline acquires information, extracts changes with language models, resolves them against existing entities and continuously refreshes the graph, achieving a claimed match rate above 96%.

Click to enlarge

It uses 119+ data sources, with the deep web as the primary source complemented by trade registries, news, maps, social media and other sources. Every delivered attribute includes the normalized result, a confidence score and its source trail, making the information traceable rather than simply AI-generated.

Click to enlarge

The infrastructure behind this operation is significant. Veridion runs about 3,200 Zen 5 processing cores plus 5,000 crawling cores, approximately 5.3TB of GPU VRAM across around 100 servers, and models ranging from 9B to 27B parameters. Storage is tiered between roughly 100TB of hot NVMe object storage, 640TB of warm HDD object storage and 4PB of cold HDFS capacity, complemented by a 252TB Cassandra database capable of millions of writes per second. The infrastructure spans three replicated data centers, with the core data plane located in Germany.

Click to enlarge

Customers consume the same knowledge graph through six APIs, Match, Enrich, Search, Location, ESG and Corporate Groups, or via CSV, JSON and Parquet batch feeds. Veridion also supports native data sharing through Snowflake, BigQuery and Databricks, while an MCP interface is coming soon, allowing AI agents to directly interrogate the graph. API match-and-enrich latency is around 1.5 seconds and below 200ms for cached records.

Commercially, Veridion sells data access rather than user seats, mainly through annual enterprise contracts, complemented by project engagements, embedded/reseller relationships and data marketplaces. Its target markets include credit agencies, insurers, financial-data providers, private equity, procurement, supply-chain and third-party-risk platforms. The overall proposition is straightforward: replace periodically refreshed, registry-centric company databases with a live, machine-built and evidence-backed representation of the global business landscape, increasingly designed to serve not only traditional enterprise applications but also AI agents.

The image below illustrates some information you can generate from the data collected, normalized and cleaned by Veridion.

Click to enlarge