Making Complex Systems Understandable with AI

I build bespoke AI applications for companies whose problems can't be solved with off-the-shelf tools.

Not chatbots. Not demos. Systems that integrate deeply with your distributed infrastructure and deliver real value.

For technical teams with problems worth $150K+ to solve.


The Problems I Solve

You Can't See What Your Systems Are Doing

Your distributed infrastructure generates data streams you can't fully understand. I build applications that make complex architectures explorable through real-time visualization, intelligent querying, and pattern detection.

Your AI Is Either Too Dumb or Too Autonomous

Most AI is either useless chatbots or automation that makes decisions humans should control. I design human-in-loop systems where AI suggests and analyzes, humans decide and refine, and both learn from the interaction.

Your AI Forgets Everything

Generic AI assistants start from zero every conversation. I build project-scoped memory systems that maintain context across interactions, with retrieval pipelines that surface the right information at the right time.

Your Internal Services Aren't AI-Accessible

I create Model Context Protocol (MCP) servers that connect AI assistants directly to your infrastructure. Your databases, APIs, and internal tools become available to AI within conversations, with proper access control.


Ideal Client Profile

You have:

  • AI integration challenges that off-the-shelf tools can't solve
  • Complex systems (graphs, distributed architectures, real-time streams) needing better visibility
  • Struggle with balancing AI autonomy and user control (not chatbots, not full automation)
  • Internal services that should be AI-accessible (MCP servers, retrieval pipelines)
  • Technical teams who value shipping over process
  • Budget for specialized work ($150K+ project scope)

Target sectors: Environmental monitoring and geospatial · Engineering design tools · Industrial IoT and sensors · Infrastructure and observability · Supply chain and logistics · Scientific computing and research


What I Don't Build

  • Generic chatbots or "talk to your data" demos
  • One-size-fits-all SaaS products
  • Impressive demos that don't deliver value
  • Solutions where existing tools would work

If your problem can be solved with off-the-shelf tools, I'll tell you that.


How We Work Together

I work with 2-3 long-term clients on retainer or project-based engagements. Deep relationships, specific problems, measurable outcomes.

  • Retainer relationships: $10-15K/month for ongoing development
  • Project contracts: $150K+ for 6-12 month builds
  • Pilot projects: $25K+ to prove value on one specific problem

Working style: Remote-first, outcome-driven, pragmatic


Technical Approach

React/TypeScript frontend engineering combined with AI systems integration and distributed systems architecture. I build unified interfaces for complex technical infrastructure — not just pretty frontends, not just AI wrappers, but integrated systems that work.

Stack: React/TypeScript · Real-time data handling · LLM orchestration · Geospatial and graph/network visualization · MCP server development

The technical choices depend on your problem. I use what works, not what's fashionable.


Examples

Building with MCP — Connecting AI assistants to real infrastructure

Frontend Maximalism and LLM Context — Architectural patterns for LLM integration in React applications

Example Projects — Custom AI applications and React engineering work


Common Questions

Can't we just use ChatGPT/Claude/other existing tools? If your problem can be solved with existing tools, I'll tell you that. I work on problems where off-the-shelf doesn't cut it.

Why these price points? Bespoke AI integration requires rare expertise. You're paying for specialized knowledge that saves you months of trial-and-error and delivers systems that actually work.

Can you do a smaller project first? Yes — $25K pilots are designed to prove value before larger commitments.


Bottom Line

You're here because generic AI tools aren't solving your problem. You need something built specifically for your systems, your data, your workflows.

I build exactly that.

Email: info@orbital.co.nz


Latest Articles

MCP Client Feature Support

Detailed comparison of Model Context Protocol (MCP) feature support across leading clients — including Claude Desktop, Nanobot, Cline, Cursor, Windsurf, and Postman — examining how each handles tools, resources, prompts, sampling, OAuth, and other key capabilities.