Data products built from the public record.
Vepisk turns difficult public sources — hundreds of small-agency websites, meeting packets, scanned minutes — into structured intelligence that people can act on.
- 7.CONSENT CALENDAR
- 7.bApprove minutes of the regular meeting
- 7.cReceive and file monthly investment report
- 7.dAdopt Resolution 26-14 approving the Five-Year Capital Improvement Plan, FY 2027–2031CIP adoption
- 8.NEW BUSINESS
- 8.bAward of construction contract — Manhole Rehabilitation Project, Phase II ($412,300)construction award
- 8.cAuthorize as-needed professional services agreements for geotechnical and materials testingon-call PSA · geotech
- 9.REPORTS
- 9.aGeneral Manager's report
The work is in the sources.
The valuable public record isn't on a portal with an API. It's spread across thousands of small websites, in PDF packets and scanned minutes, published on no schedule and in no standard format. That's exactly why it's worth reading.
Collect what's hard to collect
We build pipelines that read hostile sources continuously — one-off CMSes, meeting portals, packet PDFs, documents that need OCR — and do it politely, at rates small public servers can handle.
Extract structure, not keywords
Each project gets a domain-specific taxonomy — who was awarded what, at which stage, funded how — extracted with language models and grounded in the source document, not a keyword match.
Verify before shipping
Every extracted item traces back to the public document it came from. Counts are measured, not modeled; where we estimate, we say so.
Current focus: California's public water and wastewater agencies.
We're building a pre-RFP intelligence pipeline over the board agendas, packets, and minutes of California's small public water and wastewater agencies — the districts that large procurement platforms treat as noise, and where firms still learn about upcoming work by reading agendas one at a time. The pipeline extracts contract awards, on-call lists, capital plan adoptions, rate studies, and the firms behind them, months before anything reaches a bid portal.
| 245 | public water and wastewater agencies on the roster |
| 24 months | of board meetings read per agency |
| 6,800+ | agendas, packets, and minutes processed to date |
| 5,800+ | firm mentions extracted and classified by segment |
If your firm pursues work with California water agencies — engineering, construction management, environmental, SCADA, rate and financial consulting — we're speaking with a small number of firms about what earlier signal would change for them.
Email usHow we work
- Primary sources only. Everything we publish traces to a public document; nothing rests on a model's memory.
- Small servers, light touch. We identify ourselves, rate-limit every crawl, and never hammer a public agency's website.
- Models are tools. Language models do the reading at scale; the taxonomy, the checks, and the judgment are ours.
- Public records, used respectfully. We work with what agencies publish for the public, and nothing more.