Johan Steenkamp
Summary
I help businesses build AI systems grounded in their own expertise — not generic models that could belong to any company in their industry.
With over a decade of building production systems across fraud detection, industrial monitoring, scientific computing, and geospatial intelligence, I've seen the same pattern everywhere: the most valuable intelligence in a business lives in people's heads, informal processes, and institutional memory. Most AI has no access to any of it.
Orbital is my practice. The model supplies intelligence; your firm supplies the expertise, captured and tested against real work. I capture that expertise, write it down, and build the system on the Document Substrate, a foundation that reads client documents accurately, keeps personal details away from the AI model, and checks every figure before a person signs. The result is AI that reflects your team's judgment and standards — not generic tools your competitors can license off the shelf.
Principles
Start from the business, not the technology. Every engagement begins by understanding what your organisation actually knows. The technology follows the problem, not the other way around.
Amplify people, don't replace them. The goal is to make your best people's judgment, patterns, and instincts available everywhere they're needed — not to automate the people who developed them.
Keep the model out of the parts that have to be defensible. The model reads and suggests. Code checks and decides. A person signs. Personal details are removed before the model sees anything.
Build to last, not to depend. Systems should keep working and improving after the engagement ends. Your team should be able to maintain and extend the intelligence base themselves.
Ship products first, build infrastructure later. I focus on solving specific problems with working code, not building elaborate frameworks or perfect architectures.
One person who ships beats a coordinated team. You get me, not a team of juniors. The builder archetype matters more than ever — own the full stack from problem to solution.
How I Work
I work in three phases — Map, Architect, Build — starting from the expertise already in your firm and shaping the AI system around it.
Map: Short sessions with the people who do the work, starting from the workpapers, checklists and letters they already produce and one real job followed end to end. You receive a practice manual seed in the firm's own words, a completeness specification for each job, and an architecture note. Paid for on its own, and yours whether or not we build.
Architect: The architecture note becomes a design: which of your document types, checks, workflow rules and manual go onto the Document Substrate, which general parts already exist, and what has to be built for you. One install per customer, in accounts you own.
Build: The parts that are yours, built on the Document Substrate and managed by Orbital, so you pay only for what is yours. Every correction a reviewer makes is recorded with its reason and becomes a test case, and the checks, manual and document types are iterated as real jobs show what the first version missed.
Technical Depth
When it's time to build, I bring deep expertise in the systems that make a firm's expertise usable by AI:
- Document AI on the Document Substrate — Parsing and layout reconstruction, PII redaction with an encrypted entity vault, schema-driven extraction with every figure traced to its page, and a deterministic verification engine so a cloud model can work on client files it is never allowed to see
- Knowledge capture and intelligence architecture — Surfacing how a firm actually does the job and writing it down as completeness specifications, practice manuals and checks that run in code, so the expertise is accessible to both people and AI
- Production AI systems — LLM orchestration, MCP servers, evaluation frameworks, prompt versioning, observability, and human-in-the-loop patterns using AI SDK, Next.js, and modern AI tooling
- React applications and data visualization — Interactive network graphs, real-time dashboards, and domain-specific tools built with React, TypeScript, AntV/G6, Mapbox, and Deck.gl
- Geospatial data fusion — Multi-source intelligence overlays, temporal-spatial analysis, and interactive map interfaces for complex real-world systems
Experience
AI Product Engineer
Orbital | Self-employed Apr 2025 – Present
Building AI systems grounded in a firm's own expertise — from capturing how the work is done through to production deployment. Built and deployed the Document Substrate, a document AI foundation for professional practices, in use at a New Zealand accounting practice. Document AI, knowledge capture, evaluation frameworks, MCP servers, React applications, and geospatial intelligence.
Principal Frontend Engineer
Darwinium | Full-time Sep 2022 – Apr 2025
Led development of the fraud detection frontend — UI, dashboards, and data visualization for a platform that processes billions of digital interactions. Shaped engineering practices for modern React development. TypeScript, React, GraphQL, AntD, AntV, Mapbox.
Digital Architect & Head of Software
Syft Technologies | Full-time May 2018 – Sep 2022
Responsible for end-to-end digital infrastructure and strategy for a scientific instruments company. Led and grew engineering teams. Architected platform systems on AWS. Active developer in full-stack applications using React, GraphQL, and Next.js.
Principal Engineer
Wynyard, Cognevo, & Telstra | Full-time Aug 2014 – Apr 2018
Principal Engineer and Application Architect across multiple technology companies, focusing on data visualization frontends for security products. Led adoption of GraphQL and modern React patterns. Node.js, React, Redux, GraphQL.
Skills
- AI: AI SDK, LLM APIs, Model Context Protocol (MCP), Prompt/Context Engineering, Evaluation Frameworks (Evalite), Observability (Langfuse, OpenTelemetry), Hybrid Retrieval (pgvector, Reciprocal Rank Fusion)
- Document AI: Document parsing and OCR (LlamaParse, LlamaExtract), PII redaction and entity vaults, schema-driven extraction, grounded figures, deterministic verification, durable pipelines (Inngest)
- Knowledge Capture: Completeness specifications, practice manuals, checks that run in code, correction records as test cases
- Frontend: React, Next.js, TypeScript, Tailwind CSS, shadcn/ui, GraphQL, Mapbox, Deck.gl, AntD, AntV (G2, G6)
- Geospatial: Multi-source data fusion, temporal-spatial analysis, network graph overlays, interactive map intelligence
- Backend: Node.js, AWS, DynamoDB, S3, Postgres, SQLite, Supabase (database, auth, storage), Clerk
Education
BSc Eng (Electronics) University of Cape Town
Example Projects
Examples of AI systems and applications I've built:
- Document Substrate: Document AI for professional practices — parse, PII redaction, schema-driven extraction and a deterministic verification engine, so a cloud LLM can work on client files it is never allowed to see. Deployed for a New Zealand accounting practice
- Cycling Training Intelligence: Interactive 3D geospatial route viewer plus a grounded AI coaching layer with three-tier memory, hybrid retrieval, a reinforcing evals loop, and an authenticated MCP server. Live at routebook.app
- Deep Search: Production-grade search app with evals, observability, and prompt management
- Darwinium Frontend: Fraud detection UI with interactive dashboards and network graphs
- Syft Instrument Health: Mass Spectrometry instrument performance monitoring
- Mersen R-TOOLS MAXX: Heat sink design and thermal simulation tool