Peregrine and the quiet rewiring of public safety.

Who they are, in May 2026

Walk into the operations room of a mid-sized American police department on a Tuesday afternoon and you will probably find a sergeant staring at four monitors. One shows the records management system. One shows computer-aided dispatch. One shows the license-plate-reader feed. One shows a spreadsheet that, for reasons no one alive can explain, lives on a shared drive named "FINAL_v3."

Peregrine is the company arguing that this should not be the job. That a sergeant's job is judgment, not tab-switching. The pitch is unglamorous on paper and slightly thrilling in practice - take every disconnected system an agency owns, fuse it into one ontology, and put a single search bar on top.

By May 2026 the company is FedRAMP High authorized, valued at $2.5 billion, staffed by around 400 people, and embedded inside more than a hundred public safety agencies from Atlanta to Edmonton. The CEO still answers his own emails.

The problem they saw

American public safety runs on a strange truth - agencies are simultaneously drowning in data and starved of it. A typical municipal police force operates between fifteen and forty distinct software systems, most of which were procured in different decades, by different administrations, with different theories of how a database should behave. The result is a kind of organized amnesia. The information exists; the cop on shift cannot get to it.

Nick Noone, who would later co-found Peregrine, watched this play out for years inside Palantir's U.S. Special Operations unit. The intelligence community, for all its flaws, had budgets large enough to paper over interoperability problems. The Atlanta Police Department does not.

And so the question that became Peregrine was almost insultingly basic - what if a beat officer, a dispatcher, and a homicide detective could all search the same unified record in plain English, and get an answer in under a second? The honest answer was: nobody had bothered to make it work on a local government budget. So Noone, joined by Ben Rudolph - a UN Refugee Agency technologist and former Stanford gymnast, an unusual résumé combination by any standard - decided to.

The founders' bet

In 2018 they made what looked, to most venture capitalists at the time, like a bad wager. Selling enterprise software to municipal government is famously slow. Procurement cycles run twelve to eighteen months. Champions get reassigned. RFPs are written to be unanswerable. The category had eaten promising companies before.

Peregrine's bet was that the bottleneck was not really government's patience - it was the software's quality. Build something a cop genuinely wants to use, and the procurement department will catch up. Forward-deploy engineers into customer agencies. Deploy in twelve weeks, not twelve months. Charge less than the incumbents, then deliver more.

It is the same playbook Palantir wrote, run on cheaper rails, aimed at the people most enterprise software ignores entirely.

The product, plainly described

Peregrine is, at its dullest, a data platform. It ingests records from the agency's existing systems - records management, dispatch, body-worn cameras, automated license-plate readers, jail intake, even the occasional unloved spreadsheet - and stitches them together using an entity-resolution layer that knows that "John Q. Smith," "Smith, J.," and that license plate from last March are likely the same person.

On top of that sits the part users actually touch - a search bar, a map, a network graph, a dashboard, an alerting system. Newer to the stack is Peregrine AI, which lets analysts ask natural-language questions of their own data without writing SQL. There is also Trust at Peregrine, a permissions and audit layer that, by design, makes every query traceable to a badge number.

The proof

The case for Peregrine is, ultimately, made by who has actually signed contracts. Los Angeles Police Department. San Francisco PD. Atlanta. Miami. Fairfax County. Loudoun County Sheriff's Office. Glendale. Anaheim. Edmonton, north of the border. Peel Regional, also north of the border. Ada County Paramedics, for a change of pace. Northern Arizona University, because campus public safety has the same data problem in miniature.

Investors include Sequoia, who led the Series C, plus Founders Fund, who has been there since the beginning. The platform itself runs on a stack engineers actually like to talk about - Python and Django for application logic, Kafka and Airflow for ingestion, Postgres and Elasticsearch for retrieval, Neo4j for the graph layer, and AWS GovCloud beneath it all. AI workloads run on AWS Bedrock and, increasingly, Claude.

The mission, said carefully

Public safety technology is a category with a justifiably wary audience. The history of selling software to police is dotted with overpromises, civil-liberties scandals, and PR departments that confused "interoperability" with "surveillance." Peregrine's positioning is conscious about this. The pitch is about helping agencies use the data they already legally collect, not collecting more. The audit layer is built into the platform, not bolted on. Governance is treated as a feature, not a compliance tax.

Whether that distinction holds in practice is the conversation the company invites - and the one it will spend the next decade earning the right to have.

Why it matters tomorrow

If Peregrine's bet pays off, the second-order effect is more interesting than the first. The first-order effect is that policing gets faster and arguably more accurate. The second-order effect is that the digital plumbing of local government - the actually-difficult layer where fire, EMS, emergency management, courts and city hall do or don't talk to each other - finally has a vendor who treats it as a software problem instead of a procurement problem.

That is the bigger market, and Peregrine's expansion into fire-rescue, EMS, corrections, and emergency management suggests they know it. The Canada launch suggests the model is exportable. The FedRAMP High authorization suggests the federal door is now open. A company that started by helping a beat cop find an incident report is, four years later, plausibly building the operating system of public-facing government.

Back to the operations room

Return now to that sergeant on a Tuesday afternoon. In an agency running on Peregrine, the four monitors collapse into one. The spreadsheet named FINAL_v3 still exists somewhere - it always will - but it is ingested, parsed, indexed, and joined to everything else. The sergeant types a name. Three records, two priors, one open warrant, a flagged plate from the morning. Two seconds.

The job becomes judgment again. Which, when you think about what we hire sergeants for, was always the point.