Why we built a new kind of pipeline for alternative company data — and how it helps you focus on higher-value problems

SEC filings are supposed to tell you everything you need to know about a public company — but they’re not designed for machines. Between inconsistent formats, ambiguous sections, and the complete lack of structure outside of XBRL tags, extracting meaningful information at scale is still a painful, manual process.

There are great tools for pulling structured financials from XBRL or iXBRL, and others that can extract addresses and tables — but what about the stuff that doesn’t fit into those schemas?

• Who’s the CEO, really?
• What does the company actually do, in plain English?
• Is there a business summary we can trust?
• Do we know how many people they employ?

These aren't just "nice to have" fields — they’re critical for decision-making, yet they typically require brittle scraping, manual review, or just accepting missing values.

At ViaNexus, that’s the problem we set out to solve.

The Gap: Alternative Data That’s Still Unstructured

There’s plenty of innovation around financial statement ingestion — and rightly so. But the “metadata layer” that surrounds a company’s core identity, leadership, and operations has mostly been ignored.

This information lives in raw HTML filings — not in XBRL. It’s inconsistent, fragmented, and hard to extract even for humans. But when cleaned up, it’s incredibly valuable for downstream systems: AI agents, dashboards, LLM summarizers, portfolio screeners, and more.

That’s why we built the SEC Company Descriptions pipeline: to transform overlooked metadata into structured, verifiable insight you can actually use.

The Pipeline: Designed for Today, Built for What’s Next

The pipeline runs monthly, processing thousands of US company filings to produce clean, consistent, and structured outputs. It uses resilient logic to handle messy HTML, applies intelligent filtering to ensure completeness, and integrates tightly scoped LLM prompts to summarize content without hallucination.

What makes it powerful isn’t just what it extracts — it’s how it’s built:

• Flexible parsing recovers from inconsistent formats
• Every field is logged and validated independently
• Gemini only sees scoped, cleaned content — reducing token usage by 95%+
• Alerts and coverage checks ensure issues never go unnoticed

It’s not agentic — yet — but it’s designed for that future. Structured inputs, minimal guesswork, and full traceability make this pipeline an ideal foundation for AI-driven workflows and automated intelligence.

Built-In Quality, So You Can Focus on What Matters

We built this pipeline so you don’t have to. It’s not a framework, a toolkit, or a how-to guide, it’s a fully managed data layer designed to slot directly into your systems.

Every layer — from extraction to QA — is engineered to deliver clean, reliable inputs straight into your models, dashboards, and agents. That means no brittle scrapers, no hallucinations, and no wasted cycles patching holes in your pipeline. Our job is to deliver structured insight you can trust. Quality isn’t an afterthought — it’s built into the architecture. This frees your team — and your stack — to focus on what actually matters: surfacing opportunities, generating alpha, and making decisions that move the needle.

We handle the plumbing. You capture the upside.

How Customers Are Using It

The SEC Company Descriptions dataset supports a wide range of use cases across finance, AI, and data infrastructure:

• Monitor executive changes and track business evolution using structured metadata
• Feed clean summaries into dashboards, research platforms, and LLM assistants
• Power alerting systems for leadership shifts or filing activity
• Reduce LLM token usage by isolating only the relevant content
• Enrich internal systems with standardized company context, linked to your existing data

Whether you're screening tickers, building chat-based tools, or maintaining knowledge graphs — this dataset helps you move faster with cleaner, more reliable inputs.

One of Many: Part of Our CORE Data Bundle

Company Descriptions is just one of over 20 datasets included in our CORE data bundle at ViaNexus — a foundational set of normalized, production-ready financial data that powers everything from dashboards to AI workflows.

While most datasets focus on prices, fundamentals, or corporate actions, Company Descriptions adds the missing layer: structured context around what a company actually does, who runs it, and how it describes itself. This makes it a natural complement to time-series and event-driven data — and a key enabler for smarter filters, alerts, and LLM prompts.

Other datasets in the bundle include:

• Historical prices
• Corporate actions
• Reference and symbology data
• Filing metadata…and more — all cleaned, linked, and continuously updated.

We use CORE’s own Symbols Reference Data to kickstart this pipeline — mapping active tickers to CIKs for SEC querying — and customers can access the same building blocks. For those looking to replicate this setup, our As Reported SEC Filings dataset offers full access to raw SEC filings enabling direct integration with unstructured 10K, 10Q, and 8K content. And if there’s a specific field or dataset you’re looking for, just reach out — we’re always looking to onboard more and support your workflow.

“You don’t need to teach an agent to read the whole filing — you just need to give it what matters.”

If you're working with SEC data, building AI infrastructure, or looking for trustworthy alternative data — we’d love to connect.

🔗 Learn more: https://vianexus.com/

📧 Contact us: support@vianexus.com