How an AI startup learned to read 80 million pages of city hall paperwork.
Selling to the government has always been a peculiar sport. The buyer is rich, the buyer is real, and yet the buyer is impossible to find. Procurement signals hide inside PDFs no one indexes. Budgets are voted on in Tuesday-night meetings nobody attends. RFPs appear with two-week deadlines that favor whoever was already in the room. For decades, the response from sales teams has been to hire a former municipal employee, buy a database from the 1990s, and pray.
NationGraph is the response from the AI generation. Founded in 2024 by Kimia Hamidi and Eden Ding, the company indexes more than 110,000 state, local, and education agencies, ingests over 4 million government webpages, and continuously reads the boring documents - council minutes, capital improvement plans, contracts about to expire - that actually predict what a public agency will buy next.
The pitch is uncomplicated. Vendors who know about a contract before it becomes an RFP win the contract. NationGraph is the tool that tells them.
What's in the index.
Three tools, one job.
NationGraph's platform is built around a single workflow: notice a buying signal, research the buyer, run the play. Each piece of the product maps to a step a salesperson used to do by hand.
The radar
An AI layer that cross-references meeting minutes, budgets, purchase orders, contracts, and capital plans. When a school district approves a $400K curriculum line, the right vendor finds out before the RFP exists.
The outreach
Ranked accounts, AI-drafted emails, call scripts tailored to a specific procurement officer, and CRM sync. Built to act on signals at the speed they arrive, not the speed a quarterly review allows.
The roster
Verified emails and direct phone numbers for decision-makers across more than 110,000 agencies, with zero-bounce validation. The boring layer that makes the rest of the platform actually usable.
Co-founders.
Kimia Hamidi
University of Victoria. Previously founded price-intelligence startup Buyer (acquired by Ramp in 2021), then ran savings at Ramp. Started NationGraph to apply the same playbook to the public sector.
Eden Ding
University of British Columbia. Former quantitative developer at Citadel. Leads the engineering team that ingests, parses, and ranks the document firehose powering NationGraph's signals layer.
Why now, and why this.
Public sector procurement is a market of staggering size and stubborn opacity. The state, local, and education segment alone moves something close to $2 trillion in annual spend across hundreds of thousands of buyers. None of those buyers act like an enterprise sales target. They publish meeting agendas in HTML from 2003. They post budget books as 400-page PDFs. They sign contracts that quietly auto-renew unless someone on the vendor side happens to notice on Tuesday.
The old way to compete here was to hire former government employees and lean on relationships. That worked when the market was small and predictable. It does not scale. NationGraph's bet is that large language models, finally good enough at reading messy semi-structured documents, can do at machine speed what a procurement consultant used to do at the speed of coffee meetings.
The company's customers reflect how wide that opportunity actually is. NationGraph's roster includes enterprise software providers selling to school districts, alongside janitorial supply companies selling brooms to county facilities. Both are buying the same product for the same reason - the public sector buyer was unreachable until somebody read all the paperwork.
The Series A, led by Menlo Ventures in early 2026, was joined by the Perplexity Fund and existing investors XYZ Venture Capital and Reach Capital. The money is going where you would expect: more data coverage, more automation surface, more sellers on the platform. Toronto continues to handle most of the engineering build. San Francisco handles the customer-facing motion. Miami is the wild card and the time zone.
The short history.
Five things you didn't know.
- CEO Kimia Hamidi's previous startup, Buyer, was acquired by Ramp in 2021.
- CTO Eden Ding spent his pre-founder years writing quant code at Citadel.
- About half the team is based in Toronto, despite the SF HQ.
- The Perplexity Fund - tied to the AI search company - is on the cap table.
- Customers range from enterprise SaaS to janitorial supply. Same buyer, different product.
Where to find them.
Back to the county clerk.
That budget amendment we mentioned at the top - the one being filed right now in some county office - used to disappear into a PDF on a website nobody bookmarks. The vendor who could have helped, the one whose product happens to solve exactly that line item, would have learned about it nine months later, when the contract was already signed with somebody else.
Now the document gets read by an AI before the meeting ends. Tomorrow morning, a sales rep in a different time zone opens a CRM, sees a fresh signal, and writes one email. The black box is still a black box for most of the market. NationGraph is the company quietly handing out flashlights.