MJ Bonanno

Public safety infrastructure Live in production

New Jersey law enforcement runs on software I architected. So do agencies in eight more states.

State police, county, municipal, and corrections. The FBI and Homeland Security in New Jersey are on it too.

674
agencies
9
states
500,000
notifications a day, agency to agency

Its disability safety program, NPS-AID, has participants in more than 40 states. Built with zero venture capital.

MJ Bonanno

I’m MJ Bonanno. I build technology for the industries software left behind: law enforcement, disability safety, union benefits. Twenty years of systems that aren’t allowed to fail, and the AI infrastructure underneath them.

Counts are audited; positions within each state are illustrative. Pulses run agency to agency, the way OPS LENS moves alerts, at the real daily aggregate of about 6 a second. Simulated at that rate, not a live feed.

Book me to speak See the work

The work

five platforms in production.

Across four industries. Every one of them live, carrying real users, built without venture capital. Two of them below, all five on the ventures page.

01

Public safety Present

Online Policing Solutions

Founding member · Architect of all three platforms

Three platforms on one architecture: OPS LENS for real-time cross-jurisdictional alerts, NPS-AID protecting disabled and vulnerable people during encounters, and OPS Hubs for multi-agency operations. I designed the architecture from day one and I run the infrastructure it operates on. It runs across New Jersey law enforcement at every level, state police through municipal through corrections, and at 674 agencies across nine states. NPS-AID has participants in more than 40 states; its first CAD and 911 integrated deployment is live in Cumberland County, New Jersey.

onlinepolicingsolutions.com
Online Policing Solutions interface, view 1 of 3
Online Policing Solutions interface, view 2 of 3
Online Policing Solutions interface, view 3 of 3
02

AI infrastructure 2026

MuninnDB

Sole author

A cognitive database. Not a vector store, not a graph DB, not a RAG wrapper. Ebbinghaus decay, Hebbian learning, and Bayesian confidence are engine-native primitives, so memories strengthen, weaken, and evolve on their own. MCP-native, written in Go, ships as a single binary. Open source under BSL 1.1, with a provisional filed on the memory primitives so it can stay that way. I am its sole author.

github.com/scrypster/muninndb

Concepts and Hebbian links. A flare is a recall spreading activation; the fade is Ebbinghaus decay.

What I’m hired for

Three problems worth paying someone senior for.

  1. AI that survives production

    Architecture review · Build engagement

    Most AI work dies between the prototype and the deploy. I build the infrastructure underneath it: memory, retrieval, evaluation, and the unglamorous plumbing that decides whether a feature ships or quietly gets shelved.

    Built MuninnDB, a cognitive database, and Huginn, a local agent platform.

    See the code

  2. Legacy systems, brought forward

    Modernization consulting

    Twenty-year-old software holding up a real business. I have modernized stacks across education, healthcare, benefits, retail, and law enforcement without the rip-and-replace fantasy that gets consultants fired.

    Fourteen years across education, healthcare, benefits, retail, and law enforcement.

  3. Senior technical judgment

    Fractional CTO

    Architecture decisions, hiring, and the calls that are expensive to get wrong. I scaled an engineering organization from five to twenty-five across distributed teams and led it through an acquisition.

    Director of Engineering within 30 days of joining as a developer. Scaled that org 5 → 25 and through an acquisition.

Get in touch

Ask me the thing
you’re stuck on.

Not a brief. Not a project scope. A question about your architecture, your stack, your hiring, or whether the thing you’re about to build survives production. Half-formed is welcome.

  • Two or three sentences is plenty.
  • Fractional CTO, consulting, or a one-off second opinion.
  • You’re writing to me directly, not to a firm.
  • Prefer LinkedIn? Message me there instead.