Know exactly what AI should do in your business, and what it shouldn't. That's one session away.

For teams that know AI matters but don't have an AI department. One focused session with the specialist who owns your problem, current to today.

If you don't leave with a clear next step, the session is free.

3 specialists

One call. A team thinking behind it.

$1,327+/mo

Tooling & hardware we carry, not you

Last 48 hrs

How current every answer you get is

Claude OpusGPT-5GeminiLlamaMistralSAM2WhisperMLXCUDAvLLMLoRA fine-tuningRAG pipelinesTool-using agentsOpenClawHermes AgentMCPComputer visionOn-device inferenceMac Studio clusterGPU workstation
Claude OpusGPT-5GeminiLlamaMistralSAM2WhisperMLXCUDAvLLMLoRA fine-tuningRAG pipelinesTool-using agentsOpenClawHermes AgentMCPComputer visionOn-device inferenceMac Studio clusterGPU workstation

Proof of work

Built it. Shipped it. Still running.

A selection of what we have built and run, by industry. Client specifics stay confidential. The work speaks anyway.

Energy · Oil & Gas

Computer vision on producing well sites

Detection models watching live oil sites right now: leaks, pooling liquids, rod contamination, spark risk. The models run on edge hardware beside the cameras and feed a geospatial operations platform we built end to end. From raw drone imagery to a deployed model on site, the entire pipeline is ours.

Public sector

Applied AI for the State of North Dakota

An AI engagement delivered for the state itself. Government work means accuracy, accountability, and careful data handling are not nice-to-haves; they are the spec. We build for environments where being roughly right is not good enough.

UAS · Drone data

Engineering with KSI Data Sciences

Ongoing work with the California company behind mission-critical drone data and video systems. Where unmanned aircraft meet real-time data infrastructure, that is a room we are already in.

Before we joined forces

Samsung SmartThings

Built the first IoT app to ship on Samsung's 2017 Smart TV line.

Target HQ

Rapid prototyping for the innovation team, and trained 200+ Target engineers.

RocketSpace, San Francisco

Taught developer bootcamps at the SF tech accelerator.

The HIVE, Grand Forks

Runs the venture-backed UAS and defense accelerator in the country's drone capital.

The data engine

Training data is the unglamorous moat. We do not outsource it; we build it, by hand and with AI doing the heavy lifting, then train and deploy on our own hardware.

5,000+

images annotated by hand

100,000s

more annotated with AI assistance

Dozens

of detection classes in production pipelines

And the occasional one-off: when something needs to exist by Friday, a site, a tool, a prototype, we ship it whole. This website and the booking engine behind it were built in-house. Obviously.

The team

Three specialists, not three generalists

Each of us lives at the top of a different lane of AI work. A call routes you to the one who owns your problem. Never a guess from someone stretching outside their depth.

JR

Johnny Ryan

Founder · AI Strategy & Adoption Lead

Johnny founded B3 Consulting and leads the practice, equal parts engineer and dealmaker. He built the first IoT app to ship on Samsung's 2017 Smart TV line, rapid-prototyped for Target HQ's innovation team and trained 200+ of its engineers, and taught developer bootcamps at San Francisco's RocketSpace accelerator. Today he runs The HIVE in Grand Forks, the venture-backed UAS and defense accelerator where he's twice hosted Kevin O'Leary. At B3 he leads AI strategy and adoption, pairing hands-on AI and drone systems with the business instincts of a financial economist.

Leads

AI Strategy & AdoptionSoftware, IoT & Drone SystemsAI-Driven Marketing & Growth
Book Johnny
GE

Garett Elwood

Co-Founder · AI, Backend & Database Lead

Garett co-founded B3 Consulting after years building production ML systems: annotation pipelines, computer-vision detection models, and on-device inference. He runs the hardware lab, owns B3's backend and database work, and leads engagements where AI has to actually ship: tool-using agents, automations, and the pipelines that wire LLMs into real workflows.

Leads

Computer Vision & Data AnnotationLocal & Offline AIAI Agents & AutomationBackend & Database SystemsPrompt Engineering & Context Design
Book Garett
JH

Jacob Hanson

Co-Founder · AI Research & UI/UX Lead

Jacob co-founded B3 Consulting and keeps the team honest about what's actually working. He owns fine-tuning, retrieval-augmented generation, and model evaluation, and leads B3's UI/UX practice: interfaces people actually want to use, backed by testing rigs that separate a good demo from production quality.

Leads

RAG & Knowledge SystemsGeospatial & Orthomosaic MappingUI/UX & Product DesignModel Fine-Tuning & Evaluation
Book Jacob

Expertise

What you can actually book us for

The lanes we work in every day. Pick the one closest to your problem. The knowledge graph shows the exact tools and hardware behind each.

Current to today

TODAY IS

In AI, anything more than 48 hours old is already dated. Every call works from what is true right now, not last quarter's playbook, not a blog post from a model two versions ago.

A field-shifting AI release lands, on average, every 10.5 days. Statistically, the next one is due in:

––Days
:
––Hrs
:
––Min
:
––Sec

Whatever lands next, we will have tested it before your next call. The people who already booked hear about it first.

Keeping pace with that costs each of us, every single week:

Release notes, papers & model cards6 hrs
Hands-on testing of new models & tools8 hrs
Framework churn: migrations, new SDKs5 hrs
Community signal: forums, talks, code4 hrs
Per specialist, per week23+ hrs

More than half a working week, before any client work happens. You book the session. We carry the overhead.

[ Our free weekly brief ]

The frontier research we do anyway, yours for free.

We pay for every major AI tool, run a local-AI lab, and test what actually ships. The Frontier Brief is the free distillation: the 2 to 4 developments each week that change your costs, capabilities, or risk, and exactly what to do about each. Five minutes, no charge, unsubscribe anytime.

The investment

What it costs to know what we know.

To answer your AI questions yourself, you would need every frontier subscription, the API spend, and a hardware lab to test it all: $1,327 a month, plus half a working week keeping up. Or you book the session and skip the bill.

Frontier subscriptions

Claude Max

Anthropic

$200 / mo

ChatGPT Pro

OpenAI

$200 / mo

Google AI Suite

Google

$250 / mo

Frontier API Credits

Anthropic · OpenAI · Google

$150 / mo
Recurring subtotal$800 / mo

Hardware lab

Mac Studio Cluster

Apple · 3 × $4,000

$12,000

GPU Workstation

NVIDIA · 1 × $6,000

$6,000

Raspberry Pi Fleet

Raspberry Pi · 4 × $95

$380

+ 1 more rigs and devices

Edge hardware, drones, test devices

$600
Hardware subtotal$18,980

Our all-in monthly bill

$800 subscriptions + $527 hardware (over 36 mo)

$1,327 / mo

And the line that never shows up on an invoice: for every call we take, the three of us strategize around it beforehand, so the answer you get is the team's, tested, not one person's guess.

Behind every call

One session with one of us is really 3 minds on the problem.

You book a session with the specialist who owns your topic, but the other two are already in it. We brief each other on every engagement, so you get a team that keeps itself current, not a lone consultant.

The stack on the left is our bill, not yours. Your only cost is the session you book, and your rate is shown before you confirm.

Book a consultation

On-site workshops

Bring the whole team into the room.

For a company that wants real momentum, not a one-off call. All three of us, on site, hands-on with your tools and your people. In person only. We come to you or you come to us; travel is billed separately, at cost.

Full-Day Workshop

One full day, on site

By inquiry

scoped per engagement

The whole B3 team in the room with yours for a full day of hands-on AI work, built around your real tools and data.

  • All three of us on site for the full day
  • Hands-on with your actual tools, data, and workflows
  • A working prototype or a concrete adoption plan you keep
  • Discovery up front so the day is built around your goals
Request a quote

Two-Day Workshop

Two consecutive days, on site

By inquiry

scoped per engagement

Two days to go deeper: strategy and direction on day one, building alongside your team on day two.

  • Everything in the full-day, across two days
  • Time to actually build, not just plan
  • Your team trained on whatever we ship together
  • Discovery up front so both days are scoped to you
Request a quote

Priced per engagement. You pick a date and tell us what you need; we scope it with you and send a quote covering the work plus travel before anything is locked. No payment today.

Ready when you are

Bring the problem you've been circling. Leave knowing exactly what to do next, tested against the frontier and current to today. If you don't, the session is free. Your rate is shown the moment you pick a time.