Alex Key AI Engineer

Spatial software
is changing.

AI agents are remaking software in every industry: users describe the outcome they want, and the software gets them there. Spatial data is next.

The catch

“Isn’t there some AI for that?”

Your users have started asking. The hard part isn't adding AI - it's making it useful:

Find where an agent creates real value in your product and its workflows.

Make it work with geodata, GIS, and point clouds - not just a chat interface.

Move beyond the demo to something your customers can actually rely on.

See how I work
01 - What I do

I help you nail
this transition.

I'm an AI engineer who works with spatial software vendors as a contractor - hands-on, embedded with your team, from first experiment to production in products and platforms used by external customers.

Ⅰ

Find where agents fit

We start from your product and your users' workflows, and identify where an agent creates real value - not where the hype says it should.

Ⅱ

Built for spatial data

Agents that drive your existing tools and understand geodata, GIS, and point clouds - whether your users manage farmland, mines, or 3D scans. Deeply integrated, not a chatbot bolted on the side.

Ⅲ

Beyond the demo

Production-grade architecture, evaluation, and fine-tuning - so what ships is something your customers can rely on, not a prototype.

02 - Selected work

The future
of spatial software

Videos are hosted by Vimeo and remain blocked until you explicitly load one. Details are available in our privacy policy.

Ⅰ - tool use

An agent that drives CloudCompare

Agents are most powerful when they can reach for the tools experts already use. Here an AI agent I built drives CloudCompare to run a volume calculation for slope monitoring.

Ⅱ - panoriq

An agent for 3D point clouds

My own R&D: an agent that edits and queries point clouds in plain English - ninety seconds, no narration. The kind of agentic tooling I build for spatial-software teams.

Ⅲ - facade analysis

An agent for building facade analysis

The same agentic approach applied to imagery: the agent scans a building facade, identifies structural elements, and counts windows and objects - no manual annotation.

Ⅳ - Research Accepted · ISPRS 2026

LLM-Supervised Point Cloud Processing

From unsupervised 3D scene-graph generation to interactive scene manipulation - pairing graph-based point-cloud segmentation with an LLM agent that reasons over a scene and edits it to spec. To be presented at the XXV ISPRS Congress 2026, Toronto.

Full paper available after the congress - get in touch for a copy
Ⅴ - Course

Online course

The Spatial AI Operating System, built together with Dr. Florent Poux. Students build their own AI agent that works with spatial data - hands-on, from first principles to a working pipeline.

View the course
03 - About

Fifteen years
shipping software.

I'm an independent AI engineer based in Germany. I work hands-on - from rapid prototyping to production-oriented system design.

My focus is LLM agents and orchestration, retrieval pipelines, spatial and 3D data reasoning, and the evaluation and observability that keep them reliable. Before AI, I led software and product teams at BMW, Audi, and DriveNow, and ran my own engineering studio, Kawunu, delivering production systems for enterprise and industrial clients across mobility, manufacturing, healthcare, and logistics.

For the longer version, see my background on LinkedIn.

LangGraph LangChain LangSmith Python
04 - Work with me

Are you building spatial data software?
Let's make it Agentic

I take on a small number of engineering and R&D engagements at a time. If you build a spatial, geospatial, or 3D product for external users, get in touch.

Reach out on LinkedIn