Every business can run on AI.
Most just can't hire for it.
Kamry gives your whole office its own AI. It lives in your building, it keeps working when the internet does not, and nothing you type ever leaves.
A product of Ajian Labs Intelligence, on-premise.
Every business wants AI. Almost none can get to it.
A working AI engineer costs $2,000 to $3,000 a month, if you can find and keep one. Nine in ten African businesses report an AI skills shortage.
$2,000–3,000 per monthCloud AI is billed per seat and per token, in dollars you do not earn. A 20 person business pays $5,000 to $50,000 a year, and the bill never ends.
Forever, in dollarsEvery prompt, file and record goes to a foreign server. Client files, patient records, financials. The law has been catching up to exactly this.
45 countries · 39 regulatorsMost businesses are stuck between an AI team they cannot hire and a cloud they cannot trust with the work that matters.
We tried to hire our way into AI. We couldn't.
Before Kamry, one of us tried to build an AI model for his own logistics business. No technical co-founder, no AI background. He hit the wall every non technical founder hits: retrieval, data pipelines, memory optimisation, evaluation.
Learning it himself would take months he did not have. Hiring for it cost more than the business could justify for one project. There was no product that would simply build it for him.
That gap is Kamry.
Your knowledge base is the product. It has to stay yours.
Africa took 3% of global data-centre investment (UNCTAD 2025).
Client files cannot leave the firm's control without creating compliance exposure and professional risk.
Patient data used to power a care or triage agent faces the same cross border restrictions as any other transfer.
Verification and transaction data used to train an agent has to stay within regulatory reach.
Stop renting intelligence. Own it.
1B in + 200M out/mo
100M in + 20M out/mo
$9,450 + platform
128GB unified memory. 8 seats.
2 × 128GB domains. 20+ seats.
Paid once, in naira. No token fees. Unlimited use. Data never leaves the building, and it keeps working when the connection does not.
What's inside the box
Bought, not designed. We integrate proven silicon and open-source inference. The moat is the software and the distribution, never the chip.
What runs on 128GB
DeepSeek V4 Flash at 284B and GLM-5.2 at 753B, 2-bit, run sharded across both memory domains, scheduled overnight.
Everyone else sells you a box. We hand you a working AI.
DGX Spark at $4,699, Jetson Thor at $5,499, a DIY build, or Tiiny at $1,399. Every one of them hands you a bare computer and a command line, then leaves you to hire someone who knows what to type.
Then it runs itself
Kamry OS is built for the office manager, not the machine-learning team.
Live tokens per second and queue depth. Hot-swap models in under five seconds. Queue heavy work overnight. Seats, roles and instant revocation.
Which model and which documents produced an answer. Usage and failure rate per agent. Cloud spend avoided, in naira.
Every prompt logged on the appliance. A live statement of what left the building. Retention mapped to your jurisdiction, with verifiable deletion.
Temperature, wear and power draw. Rides out cuts and resumes after. Updates over the air or from offline media. Support sessions you switch on.
Kamry OS is specified and in active development. This round funds it, and it ships with Batch 01. We would rather say that than show you a mockup.
It also runs on hardware you already own
An AI agent, built on your business, without an AI team
Tell us what you want automated. We build the agent, grounded in your own knowledge base, mixed with public data only where it genuinely helps. You do not hire, learn or maintain any of it.
Contracts, records, playbooks, tickets. Whatever the agent needs to know.
Mixed in only where it genuinely improves the answers.
Assisted build: retrieval, fine-tuning, evaluation, done for you.
Running on your appliance, or deployed into your app and tools.
$1,600 to $2,000, scoped after onboarding, live in about a month. A fraction of one month of an AI engineer's salary.
Any business can become an AI-native business
Built right, the agent becomes the interface. The business does not have to change to become AI-native. Only the front door does.
No menu tree, no back-office re-entry. The conversation is the checkout.
Why now. Three things became true at once.
Open-source models reached frontier-class quality and kept it. The intelligence stopped being the scarce, expensive part of the problem.
Free, and improving quarterlySparse mixture-of-experts inference put 120B-class models inside a 130W envelope. On-premise stopped needing a server room or a cooling budget.
130W, on an inverterAfrican data protection laws moved from statute books into active enforcement, and cross-border transfer became a real liability rather than a policy footnote.
45 countries · 39 regulatorsAfrica's AI market grows from $4.5B to $16.5B by 2030, at 27.4% a year (Mastercard). Nigeria is second on the continent for AI startups.
Market size, bottom up
Annual addressable AI spend across Nigeria, Kenya, Ghana and South Africa.
| Segment | Organisations | Annual spend each | Total |
|---|---|---|---|
| Top law firms | 100 | $300,000 | $30.0M |
| Large enterprises in finance, telecom and industry | 50 | $150,000 | $7.5M |
| Top hospitals | 30 | $80,000 | $2.4M |
| Government data-processing agencies | 20 | $100,000 | $2.0M |
| Addressable annual spend | 200 | $41.9M |
Growing at 27% a year. Africa's AI market reaches $16.5B by 2030. Nigeria is second on the continent for AI startups, with $218M of AI venture funding in 2023.
Everyone relates. We start where the pain is sharpest.
Draft contracts, analyse case files, summarise discovery. All on the premises, all confidential.
Healthcare, then financial services.
Government, then every office renting AI.
The fastest way into a firm is somebody it already trusts
We are partnering with the companies that digitise physical records for law firms, hospitals and agencies. They finish a project and hand a client thousands of searchable documents. We turn that pile into something the client can actually ask questions of.
Document digitisation firms already sit inside the exact organisations we want, and they are trusted there.
The project ends with a client owning a large digital archive and no idea what to do with it.
The partner introduces us to the decision maker. The trust is already there and the need is obvious.
We pay a referral fee on each deal. The partner monetises a relationship twice on the same project.
Founder-led direct sales into Lagos legal, and job-board interception: companies advertising for an AI engineer have already told the market they have the problem.
How we make money
Kamry One at $9,450, Pro at $17,900. Kamry OS included and owned outright. No subscription required, ever.
$9,450 · $17,900The same software on hardware they already own. DGX, Jetson, or their own server. Not capped by manufacturing.
$100 / monthWe build the agent on their data, scoped after onboarding and live in about a month.
$1,600 – $2,000Sector packs for legal, medical, government and operations. Models priced separately.
$10 – $50 per packWe sell the box once. Everything after it recurs, and the recurring layer is not limited to customers who bought a box.
Unit economics
| Volume | Landed Lagos | Sells at | Margin |
|---|---|---|---|
| Single unit | $8,167 | $9,450 | 13.58% |
| Batch 01, 10 units | $7,740 | $8,950 | 13.52% |
| Batch 01, 100 units | $7,212 | $9,450 | 23.68% |
| OEM, 1,000 units | $6,839 | $8,360 | 18.19% |
Kamry One. Hardware ($9,450) plus $100 a month for 36 months.
Kamry One Pro ($17,900), on the same platform basis.
Delivered hardware margins (14–24%) fully absorb logistics, duties and warranty. The platform relationship compounds recurring value long after the sale.
Competition
| Stays on site | Serves an office | Ships the software | African support | |
|---|---|---|---|---|
| Cloud AIChatGPT, Claude, Gemini | ||||
| Azure, AWS regionsSouth Africa residency only | ||||
| Tiiny AI$1,399 · individuals, not offices | ||||
| NVIDIA DGX Spark$4,699 · bare dev box, FOB | ||||
| DIY buildsNeeds the talent you cannot hire | ||||
| KamryAppliance plus the operating system |
We do not win on silicon. Spark is the same memory class at a lower price. That is not the argument.
Roadmap and milestones
Team and the ask
A decade spanning mechanical, software and AI engineering, the exact convergence Kamry is built on. Co-founder and CTO of NFT Pro and Smart X. Founder of Ajian Labs with Amford.
He owns the machine: silicon, inference, thermal and mechanical design, and the path from dev board to production.
A repeat founder building AI products for African market realities. Founder of Sendrail, co-founder of Rivabit, AI product and design consultant for Afren AI and Browser GPT.
He owns product, design, operations and go-to-market. Kamry is the company his last three prepared him for.
Buys a working prototype, ten paying pilots and a production-ready v1. Seed-ready in 18 months.
1.4 billion people. The youngest population on Earth, median age 19. The fastest-growing AI market in the world, and the least served by cloud infrastructure.
Whoever builds the on-premise intelligence layer for this continent builds the next MTN.