We're building the intelligence layer small DoD contractors deserve.

Federal-contracting intelligence is fragmented across more than seventeen public data sources. Multi-INT GovCon Starter is the first product to unify them around a single question: what is my company actually qualified to win?

Mission

We're a federal-contracting intelligence platform for small-to-mid DoD contractors. The wall behind sign-up gives you capability-fit scoring against active solicitations, a knowledge-graph view of where the $850B+ DoD obligation flow goes, and saved-search alerts delivered by email.

The architecture is a knowledge-graph (Neo4j) plus vector store (Qdrant) plus an LLM swarm orchestrated by LangGraph. Every reasoning trace is hash-chained for audit. Every public-data fetch is attributed back to its source. We don't curate your watchlist — your team does, because operator judgment is still the differentiator.

V1 is UNCLASSIFIED-only. Email-only alert delivery. No customer logos on this page, because we don't have a design partner yet — we'd rather earn trust by being explicit about scope than by faking social proof.

Why we built this

If you've done capture work for a small DoD contractor, you know the pattern. The forecast PDFs live on one site, the active solicitations on another, the past-award data on a third, and the policy guidance is buried in DODI documents nobody reads. Every BD lead reinvents the same spreadsheet.

We built this because the capability-fit question deserves a real product — not a quarterly LinkedIn export. The platform handles the plumbing: graph traversal across agencies, PEOs, programs, and NAICS; vector retrieval over capability statements and solicitation text; a 6-dimensional fit-score rubric with reasoning. Your team focuses on the calls that close.

Team

Operators with DoD acquisition and AI background. We're not listing individual names in V1 — we'll add them when the design-partner program opens and the team has the customer wins to point to. Until then, the work speaks for itself.

Tech stack

For the technical buyers: the platform is built with the following components, all of them production-grade and independently auditable.

  • NVIDIA Triton
  • LangGraph swarm
  • Neo4j
  • Qdrant
  • FastAPI
  • Next.js
  • Keycloak

Triton runs the model serving. LangGraph orchestrates the agent swarm. Neo4j stores the graph; Qdrant holds the vector index. FastAPI fronts the API; Next.js drives this UI; Keycloak handles identity.

Ready to see the data?

The free explorer is open. No signup, no card.

Portolan
Market intelligence for BD, capture, and pursuit teams. By Nerdrums.
UNCLASSIFIED-only V1. Built with audit-chain provenance.