Founder-led · Milwaukee, WI · Building now

A new city for a new order of software.

Amarna is a founder-led engineering studio. We ship AI-native enterprise systems today — and we build Amarna Studio, an authoring tool for interactive e-learning with AI built in. The services fund the bet. The bet is the future.

Akhet-Aten · the horizon of the discMove · hold · release

A capital raised from empty desert in four years.
300K+
Users on training & commerce platforms we've built
523×
Production query performance gain shipped
$1M+
Annual e-commerce revenue generated
108K+
Open-source plugin downloads
25+
Years combined engineering & learning-design experience
Why "Amarna"

3,300 years ago, someone tore up the rulebook and built a city for it.

Amarna — Akhet-Aten — was the capital Pharaoh Akhenaten raised from empty desert to serve a single, radical idea: worship the Aten, the light itself, and break from a thousand years of orthodoxy. It didn’t last. But for a moment, an entire civilization reorganized around a new source of power.

AI is that moment for software. The old equation — more engineers means more output — is broken. A small, senior, AI-fluent team can now deliver what used to take a department. We're not writing decks about the transformation. We're building in it.

The Amarna wager

Old order: headcount is the lever. Scale means hiring. Quality and speed trade off against each other.

New order: judgment is the lever. Four senior builders with the right harness out-ship the department — and own every outcome end to end.

The name is the strategy: abandon the assumptions, build the new thing, name it after the ones bold enough to try.

The old order

Twenty years. A hundred thousand hands. One shape.

Khufu's pyramid is around two and a half million blocks, and every decision inside it was already made before the first course was laid. That is what the old equation buys: not a better answer, more of the same one. It was the tallest thing anyone had ever built and it held the record for three and a half thousand years — the strongest case the headcount lever will ever make for itself.

Sixty years before Amarna, this was still the whole horizon. Then one king walked eight miles into empty desert and put up a capital in four years, with a new god, a new way of drawing the human body, and a fraction of the labour. He lost. Giza is still standing. We know the odds and we would take the same bet again — because the lever moved, and pretending otherwise is how you end up building the biggest possible version of the wrong thing.

Six glyphs. Hold one and the plateau answers.

Giza · six pyramids and a lion, evaluated per pixelMove to stir the wind · scroll to set the sun · hold and release to blow the sand

The other half of the business

Nine in ten enterprise agent projects never reach production. That gap is the engagement.

Amarna is an engineering studio that takes client work. That half came first, it pays for the rest, and it is not a side door — the products on this page are what the same team builds when it points the same harness at its own problems. Every capability below is something we've shipped for a real company, on a real deadline, for real users.

We work in one shape: a fixed-scope build to an outcome we agree on up front, then a retainer that keeps it working. The retainer is not padding — an agent that shipped in March is not the same agent in September, because the model underneath it changed and nothing told you.

Enterprise agent projects that never reach production — Stanford HAI AI Index, 2026
89%
Enterprises reporting significant agent ROI — Stanford HAI AI Index, 2026
23%
Forecast AI agent software spend, 2026 — Gartner, May 2026
$206.5B
Enterprise app spend exposed to agentic disruption — Gartner, 2026
$234B

These four are third-party figures and are cited as such. Every other number on this page is ours and is checked against the code it describes.

01

AI Agents That Reach Production

The demo is not the hard part — the model does that on its own. What ships is the layer above it: tools an agent cannot misuse, a written mandate for what it may decide, gates that check the work before anyone downstream sees it, a human in the loop where being wrong costs more than being slow, and a hard cost ceiling per session.

  • Nine validation gates and a repair loop before a draft compiles — Amarna Studio
  • A written mandate per task, bounding what the agent may agree to — Amarna Agent
  • Every figure in an answer checked against a retrieved passage — Amarna CX
02

Rescue for Agents That Stalled

Pilots stall in the same four places: a tool surface the model cannot use reliably, no evals, no traces, and a model version that moved underneath it. We take the one that demoed well and did not survive contact — instrument it, build a golden set out of your real traffic, rebuild the tool layer, and put a defensible number on what it actually resolves.

  • Golden-set evals from real traffic, not synthetic prompts
  • Traces, per-session cost ceilings and model-version regressions
  • Prompt-injection and tenant-isolation review before anything is customer-facing
03

AI-Powered Product Development

Real AI in your architecture — ML models on your data, agentic workflows with proper guardrails, cost management, and fallback behavior. Not a text box calling an API.

  • ML churn prediction across 1,250+ B2B accounts
  • RFM segmentation on live transaction data
  • Agentic engineering workflows in production
04

Enterprise Modernization at AI Speed

Legacy .NET migrations accelerated 3–5× by an AI-augmented workflow. We've done the hard ones — WCF → Minimal APIs, monolith → microservices, on-prem → cloud.

  • Zero-downtime WCF → Minimal APIs via YARP
  • Azure App Services → Dockerized AWS EKS
  • Platform-wide real-time infra across 14 services
05

Intelligent Integration

Smart connections with AI-powered transformation, anomaly detection, and automated error recovery — handling real transactions for real businesses.

  • Dynamics 365 ↔ Optimizely Commerce, bidirectional
  • PingOne / Azure AD identity federation with RBAC
  • Multi-currency payments with fraud detection
06

Data → Intelligence

Years of data, zero analytics? We turn raw enterprise data into automated business intelligence — on your existing .NET / SQL Server / Azure stack, no separate data team required.

  • Zero-to-ML churn platform ("Orderly")
  • Real-time BI dashboards for business users
  • Cross-system pipelines with error recovery
How the work is bought

Three shapes, and we will tell you which one you are.

Priced against the cost of the work being replaced, not against hours. Which shape fits is a question about your data and your appetite for risk, and it is worth ten minutes on a call before anybody writes a proposal.

01

Build, then keep it working

A fixed-scope build to a defined outcome, then a monthly retainer that carries the evals, the traces, the model-version regressions and the changes the business asks for once it has met real users. The default, and the one most work should be.

  • Scope, outcome and acceptance agreed before the build starts
  • Retainer covers drift, not just uptime
  • You own the repository and the prompts from day one
02

Priced against the outcome

Where the outcome is countable and attribution is clean — a resolved ticket, a recovered underpayment, a booked appointment — we will price against it and carry some of the risk. Where attribution is muddy, we say so instead of dressing a time-and-materials engagement up as an outcome.

  • Needs a measurable event and an agreed source of truth for it
  • Baseline measured before anything is switched on
  • Ruled out honestly when the data cannot support it
03

Point it inward first

The cheapest agent to justify is one aimed at your own back office: no customer exposure, no procurement cycle, and it is measured against a cost you are already paying. Frequently the right place to start, and the fastest way to find out whether the messy part of your process can be written down at all.

  • Internal ops, claims, reconciliation, reporting, intake
  • Measured against a cost centre you can already see
  • Proves the process before it faces a customer
Shipping now · Amarna Studio

Build interactive e-learning. Publish to every format an LMS accepts.

Amarna Studio is an authoring tool for interactive e-learning. Describe the course and get a validated draft back, lay it out on a canvas, branch it into scenarios, stage it as animated scenes, and publish a self-contained package that runs in any LMS — without writing code. Run it as a desktop app, where projects are plain JSON on your own disk and nothing needs a network, or in the browser, where the same backend and the same interface sit behind a sign-in. It's built for people who outgrew slide-based authoring tools but don't want to hand-build SCORM packages.

The AI authoring layer runs on Aten, our in-house harness — the Cursor loop, inspect and prompt and regenerate and replay, pointed at training instead of code.

The model never writes the course file. It writes a small course spec — objectives, screens, decisions, feedback, scoring — that nine gates check before a deterministic compiler turns it into a real project: schema, semantics, an exhaustive walk of every path a learner can take, and objective coverage all block a bad draft and send it back through a repair loop. Attach your own policies and SME documents and it writes from those, with every citation checked against the passage it names. It runs with no API key at all, or with your own on any of eight providers.

// Cursor : code :: Aten : immersive training software

01 · Compose
Describe it, get a validated draft
One run is one undoable change: a spec, nine gates, a repair loop, then a compiled project. The keyless provider produces a real draft, so the whole loop works before you spend anything.
02 · Author
Canvas, layers, timeline
Pages, sections and layout containers with 13 element types, rich text, per-breakpoint responsive overrides, and a keyframe timeline.
03 · Branch
A scenario graph, not a slide order
A node-graph editor with 9 node types: choice points, dialogue, condition gates, variable setters, random gates, merges and endings.
04 · Assess
17 interaction types out of the box
Multiple choice through drag-and-drop, hotspot, matching, ordering and media annotation — each independently scorable and weighted.
05 · Animate
A screen, or a scene that plays
Any beat can render as a real-time animated scene on a GPU canvas — rigged characters, camera, lip sync — with flash and cut rates clamped as it compiles and a per-beat text mirror for screen readers.
06 · Check
Accessibility while you author
A built-in WCAG validator, contrast checker, tab-order manager and caption editor, so problems surface before review, not after.
07 · Publish
One project, eleven targets
HTML5, SCORM 1.2, SCORM 2004 3rd and 4th Edition, xAPI, cmi5, AICC, MP4/WebM/MOV renders, and a hosted review build — plus a public share link anyone can open without an account.
08 · Review
Stakeholders comment on the running course
Published builds hosted behind an email gate with an in-place comment overlay and live threads — feedback on the course, not a screenshot.
Publish targets
11
Interaction types
17
Element types
13
Gates per AI draft
9
Model providers
8
Paths simulated
512
In beta

Five more products, out of the same harness.

Amarna Studio proved the loop: a model writes a small structured document, gates check it before anything downstream sees it, and a deterministic compiler turns it into the real artifact. That loop is not about e-learning — it's about never letting a model hold the thing you ship. These five point it at five different problems, and the last two show its other half: when what you ship is an answer or a score rather than a file, the gates move to the moment it goes out.

All five are in beta: moving quickly, open to early users who want a say in where they go next. Each card ends with the line saying where that one’s edge is today — on this shelf those are the sentences that matter most.

Amarna Video

Beta

Type a premise. Get an animated episode you can actually edit.

// premise → document → animated episode

Everyone else generates pixels. Amarna Video generates a document that renders. A model writes ten fields per beat — actor, framing, action, emotion, line — and a pure, seeded compiler synthesises every bone transform, keyframe, mouth shape, particle and camera move from it. Nothing is sampled at render time, so nothing drifts: rewrite one line and the cast, the sets, the blocking and the shot list come back byte-identical, with only that line's mouth track and the timing after it moved. That addressable interior is the thing generative video structurally cannot have.

  • Restyle without regenerating — 12 style packs over one document
  • 54 animation controls saved as data, so a look is a link
  • Exports mp4 to 2160p in the browser, with no API key required
Style packs
12
Animation controls
54
Writing registers
6

Runs entirely in the browser today, keyless. Export quality follows WebCodecs support and the app tells you which path it took.

Amarna Agent

Beta

An agent that makes your phone calls.

// your week → a mandate → a call that gets made

Sign in with Google and your week shows up. Point at any line of it and say what the call is for — "ask if the brakes are done" — then hand it off. Amarna dials the business, talks to a human, and comes back with a transcript, a recording and whatever it agreed to on your behalf. What bounds it is a written mandate on every task: what it may agree to, what it must not, and the facts it is allowed to give out. Outside that list it says it needs to check with you and asks for a callback. It does not improvise.

  • Reads the calendar and drafts both the task and its mandate
  • Calls businesses, never people — enforced in the schema, not a policy page
  • Every call opens by disclosing who it is for, that it is an AI, and that it is recorded
From task to dialing
1 tap
Independent business-only checks
Calls that disclose
100%

Places live calls through a provisioned number. Google Calendar is a sensitive OAuth scope, so early access is by invitation while verification completes.

Amarna Dental

Beta

Three revenue apps for the practices nobody built software for.

// three apps · one platform · run one or run all

A $1m practice leaking 8% to denials and underpayments loses $80k a year, and almost none of it is visible by eye — an underpayment posts as an ordinary write-off and the ledger still balances. Recover reads the 835 remittance advice and turns X12 adjustment codes into worked appeals. FeeIQ ranks every payer by allowed production per chair hour and solves for the assumption a drop-or-renegotiate decision actually rests on. Pulse decomposes a change in collections into contributions that sum exactly to the total. All three are administrative, so FDA-exempt, and all three read Open Dental first — no five-figure integration toll.

  • Recover — denial and underpayment recovery, with scoring you can argue with
  • FeeIQ — payer scorecards, a break-even solver and a negotiation packet
  • Pulse — benchmarking with attribution that adds up, no residual
Apps, independently runnable
3
Tests over the domain logic
61
Per month, per practice
$249+

The domain logic is real and tested, running today against a seeded synthetic practice. Live Open Dental integration and the BAAs land with the first design partners.

Amarna CX

Beta

The AI customer service department, measured honestly.

// question → grounded answer → a resolution someone checked

Every support platform sells deflection — the share of customers who stopped talking to you. It scores the ones you answered wrongly and the ones who gave up as wins, which is how a vendor shows 90% deflection against 40% real resolution. Amarna CX does not compute it anywhere. Sending an answer only claims a resolution; after the re-contact window a sweep asks the one question that settles it — did this customer come back, in this thread or any other — and marks it verified or reopened. Every rate is read off that ledger rather than a counter, because a counter can be nudged and a ledger has to be rewritten. Grounding is held to the same standard: every price, percentage and duration in an answer must appear literally in a retrieved passage, or the answer does not go out.

  • Tenant isolation is a database invariant — a query that forgets its scope returns nothing, not somebody else’s transcripts
  • Autonomy is granted one intent at a time, never by default
  • Runs end to end with no API keys at all, on built-in simulators that say so on every screen
Escalation triggers, worst case first
9
Tests, isolation included
72
Deflection metrics
0

Web chat is the channel that works today; voice, SMS and email are typed seams that refuse rather than pretend. Row-level security is inert against a Postgres superuser, so it ships a non-bypassing role and refuses to start without it.

Amarna LMS

Beta

The one LMS that knows what the course was trying to teach.

// course spec → evidence → proficiency, objective by objective

Every other LMS receives a sealed SCORM package. It can report that somebody opened it, spent 22 minutes and scored 80% — and nothing at all about what that person can now do, because the content is opaque to it. Amarna owns both sides of the seam: a course authored in Studio arrives here as a competency graph, with objectives and a quality, a score delta and the objectives it touches on every choice and every interaction. So the readout is per-objective proficiency rather than completions. Objectives with no evidence behind them are reported as exactly that and counted against the roll-up — quietly dropping them is how a report claims full coverage of a course that never assessed half of what it promised.

  • An xAPI profile, not a seventh publish target — the same package still runs in Docebo
  • Courses authored elsewhere are accepted too, and labelled imported rather than refused
  • Refusals say “not found”, never “not yours” — ids are guessable and the difference is an oracle
Tests over the domain logic
73
Tables in its own schema
10
Dependency-free domain modules
4

The furthest back of the five, and the only one with no running surface yet: the domain modules, the schema and the deposit contract are real and tested, but a deposit is validated today rather than stored. The proficiency thresholds are deliberately unresolved until the first design partner.

The founders

Four builders. One shared codebase.

Lead engineers who've shipped together in production, plus the product ownership and learning-design depth behind the training software — the same four people on the client work and on everything above.

Andrew Gunn
Co-founder · Lead Engineer

Full-stack architect and applied-AI engineer, 10+ years across enterprise .NET, React, and cloud. Built e-commerce platforms grossing $1M+/yr, ML analytics across 1,250+ accounts, and AI-driven pipelines in production. 7 years running an independent practice; 108K+ plugin downloads.

Tim Griffin
Co-founder · Lead Engineer

Distributed-systems engineer and tech lead, 6+ years of enterprise .NET at scale. Specializes in event sourcing, CQRS, cloud migrations, and real-time infrastructure. Delivered 523× performance gains and the first WCF → Minimal APIs migration at a global payments company.

Tony Broerman
Co-founder · Learning Design

Nearly a decade in e-learning and immersive training content development, building learning experiences for a world leader in de-escalation and crisis-prevention training. The domain voice behind Amarna Studio — he knows what makes training actually change behavior, not just check a box.

Ryan Diaz
Co-founder · Product

Technical product owner, a decade in B2B2C and enterprise e-commerce. Led the Optimizely rebuild for a world leader in crisis-prevention training — CRM and financial-system integrations, upgraded payments, real conversion gains — after running dealer e-commerce at a national tire distributor. A certified POPM who works in the codebase: debugs alongside the team and ships PRs into the microservices himself.

Get in touch

Let's build something worth naming a city after.

Whether you have an agent that stalled short of production, a hard system that needs shipping, or a question about where immersive training goes next — we respond fast, and we'll tell you honestly if we're the right fit.