AI systems engineered to run in production.
We design, build, integrate and deploy AI software, agents and automation for organisations that need working systems — not another prototype that stalls after the demo.
- Delivery
- End to end, in production
- Engagement
- Led by the engineer who builds it
- Based in
- South Africa, working abroad
Engineering AI where it creates real leverage.
We work on the parts of a business where AI changes throughput or cost: the workflows people repeat, the documents they read, the systems that never talk to each other.
AI Software Development
Full-stack AI-native products, integrations and intelligent product features — designed, engineered and deployed as real software.
AI Agents
Autonomous and semi-autonomous systems that carry out multi-step business workflows with tools, guardrails and review points.
AI Automation
Production workflow automation wired across the systems you already run, with monitoring and a human in the loop where it matters.
Architecture. A useful AI system reads the data you already hold, applies the rules your business actually runs on, and writes results back into the software your teams use. Anything less is a separate tool nobody opens.
One of our systems you can go and use today.
We build across industries and system types. MaxLearning is the one we can show you in full — a production AI platform we developed end to end, live, with real learners under real constraints.
MaxLearning — the click-to-career platform
Not a course library. A pipeline: test, teach, coach, assess, match, place and prove — running as one system. It carries someone from their first time touching a computer through to a verified career, in their own language, on whatever device they have.
- 300+ courses — from first computer skills through to advanced digital forensics.
- Works with no signal — full courses run offline, and lessons adapt to low data and low bandwidth.
- Assessment that cannot be faked — every exam is generated uniquely per learner, and AI use is detected.
- Certificates that verify — each carries a unique ID an employer can validate instantly.
- Outcomes measured, not assumed — follow-up at 3, 6 and 12 months to check whether people are actually working.
- 18 hours of live human tutoring daily, alongside the AI.
Those are the problems most serious AI systems hit: scattered data, integrity you can defend, users you cannot assume anything about, and results someone checks months later.
Six systems we build most often.
Chosen per project on fit, cost and the operational constraints you already live with.
AI Agents
Systems that plan, call tools and complete work end to end, with review points where they matter.
Document Intelligence
Extraction, classification and review of high-volume documents, with human sign-off built in.
RAG & Intelligent Search
Retrieval grounded in your own content and records, answering from what you actually hold.
Workflow Automation
Operational processes wired across the systems that never talk to each other.
AI Learning Platforms
Assessment, certification and progression as a system — the ground we know best.
Copilots & Conversational AI
In-product assistants and service interfaces that operate on live application state and real data.
From problem to production.
Four stages, each ending in something you can look at and judge. No stage depends on you taking the next one on faith.
Discover
Understand the business objective, the workflow behind it, the people using it and the data actually available.
Prototype
Validate the riskiest assumptions quickly, on your real data, before committing to a full build.
Build
Develop the production system, the integrations around it and the interfaces people actually use.
Run
Test accuracy, reliability, security and running cost — then support it, or hand it over documented.
Built with: TypeScript · Python · React · Next.js · Postgres · Supabase · vector databases · OpenAI, Anthropic, Google and open-weight models · AWS, Azure and Google Cloud. Selected per project, never for their own sake.
The pattern holds. The vocabulary changes.
Read the data, apply the rules, write the result back. What differs between sectors is the risk, the paperwork and what happens when it is wrong.
Education & Learning
Assessment, certification and placement systems. Where we have built the most.
Financial Services
Document review, risk workflows, reconciliation and support operations.
Healthcare
Clinical and administrative document handling with human review built in.
Professional Services
Research, drafting, review and reporting workflows across large document sets.
A model that fits the work.
Some problems need a week of clear thinking; others need engineers embedded with your team for a quarter.
Discovery & Strategy
Define the opportunity, the architecture and a realistic implementation path — workflow review, solution architecture and a delivery roadmap.
Project Build
A fixed-scope product, agent or automation build delivered end to end, with defined milestones, production deployment and documented handover.
Embedded Engineering
AI engineering capability integrated directly with your team — senior engineers in your process, shared standards, knowledge transfer as you go.
Questions we get early.
The things most teams want settled before a first call. Anything not covered here, ask directly.
How long does an AI project normally take?
Who owns the software and the code?
How do you approach security and our data?
What happens after launch?
Who will actually be doing the work?
Who we build with, and who we build for.
Serious AI work does not stand on its own. It sits behind a perimeter, on top of data that has to stay protected, inside an information environment somebody may be working against you.
Security & intelligence partners
Capabilities we bring to an engagement when the risk warrants it. Three layers, addressed separately.
Outpost24
Everything an adversary can reach from outside.
Exposure and vulnerability management: attack surface management, penetration testing as a service, threat intelligence, credential-breach monitoring and identity security. Discover, test, prioritise, fix — continuously.
outpost24.co.zaCanaryBit
The data itself, protected in use.
Infrastructure and AI security: confidential computing, confidential AI, sovereign data residency and production AI implementation — so sensitive workloads stay protected even from the operator running them.
canarybit.co.zaBlack Cube
What is being said — and who is manufacturing it.
DSM-SOC: 24×7 graded intelligence on the public information environment around a principal, ministry or national programme; lawful measured counter-narrative response; and, once the evidence justifies it, forecasting, synthetic-media detection and influence-network mapping.
blackcubedefence.co.zaThe operating group
Five businesses across investment, energy, education, security and advisory. The software and AI systems behind them are designed and built by us.
Maximum Notion
Human threat intelligence and insider threat prevention — enquiry, investigation and monitoring where the risk sits with people rather than systems.
maximumnotion.co.zaMaxLearning
The click-to-career learning platform — 300+ courses, assessment that cannot be faked and placement tracking. Covered in full in our work.
maxlearning.co.zaThird Option
A private consultancy taking corporate and institutional mandates — corporate refocusing, crisis management and situations where conventional approaches have already failed.
thirdoption.aiJUA Investments
An independent power producer developing, financing, owning and operating renewable energy assets across Africa — solar PV, battery storage and hybrid systems.
juainvestments.co.zaAPEX Investments
A private investment house. Invests its own capital in businesses and infrastructure across Sub-Saharan Africa, and advises families, funds and corporates who wish to do the same.
apexinvest.ltdBring us the problem.
Tell us what you are trying to build or improve. We will tell you honestly whether AI is the right answer, and what it would take. We reply within one working day.
Every engagement is led by the person who builds it. You deal with the engineer, start to finish.