Nadine Alameh is back — former CEO of the Open Geospatial Consortium, and now CEO and co-founder of Lunate AI, a six-month-old company sitting right at the messy intersection of geospatial and AI.
In this conversation, Nadine breaks down the three types of clients she’s seeing right now: government agencies standing at the edge of the river, wondering whether to jump in, startups from outside the geospatial world stumbling in with big ideas, and organizations that know they need to modernize but don’t know who to call.
We get into why the real value today is in experience and advisory rather than raw coding, why “moving up the stack” matters more than ever, and how AI agents are quietly reshaping everything — from how satellites get tasked to how dashboards (or whatever replaces them) get built.
We also talk about the death of the one-size-fits-all dashboard, world models and simulations, why trust and guardrails are the actual hard work, and what it takes to go from a flashy proof-of-concept to something a bank can rely on every morning.
If you’re a GIS professional thinking about where to position yourself, a startup founder wandering into the geospatial world, or someone trying to figure out how AI fits into your workflows — this one’s for you.
In Conversation
Introduction: Nadine Alameh and Lunate AI
Daniel: Nadine, it’s awesome to see you again. I’ve been following your career from the outside for years — the twists and turns — and now you’ve started a new business, Lunate AI. For anyone who doesn’t know you, could you give us a quick introduction?
Nadine: Great to be back on the Mapscaping podcast. For those who don’t know me — the geospatial world is mighty but small. I’ve been in the sector for a while and seen it from various angles. Most people know me as the former CEO of the Open Geospatial Consortium, where I grew membership to its maximum size and made it truly global. We launched the AI working group, digital twins, gaming, and a startup initiative — all the exciting parts of geospatial innovation.
Nadine: I’m drawn to building or scaling things, so I’ve always been a founder, CEO, or executive director of some kind. I had my own geospatial business in aviation data management before. Now, about six months ago, three of us started Lunate AI. One partner spent her career at NASA Goddard; the other came from AWS’s geospatial division. We’re here because the whole geospatial sector is asking: what do we do with AI? What’s going to change? What won’t? We’re building a team that helps organizations navigate that modernization journey.
The Three Client Buckets
Daniel: You’re sitting at the intersection of geospatial and AI — two massive circles in the Venn diagram. What are the companies you work with actually struggling with?
Nadine: I’d break it into three buckets. The first is government organizations — federal, national, state, and local. They’ve invested in data and workflows. Now they’re asking: “What do we do with AI? Where do we even start?” This week we ran a workshop with USGS leadership. The analogy we landed on: AI is like a river that’s coming, and you’re standing on the bank wondering whether to jump. You don’t know where it goes. You can’t wait forever, but you’re afraid to commit. Our job is to help them go back to basics — why do you want to do this? What problem are you solving, before you get blinded by the shiny stuff?
Nadine: The second bucket is AI-native startups. They’re not from geospatial or earth observation, but they have big ideas. AI is lowering the barrier for everyone, and geospatial data is more accessible than ever. They stumble into our world without knowing the field, the community, or the landmines. We validate their products and architectures, connect them with the right partners — almost a go-to-market function, but really it’s product validation.
Nadine: The third bucket is organizations with a specific project ready to execute. They have a use case — border patrol monitoring, oil spill detection — and they need to know who should build it. This is where years of network become invaluable. We know the people, their customers, what those customers actually say about them, and which projects worked and which didn’t. We assemble consortiums of the best-of-breed partners and bid on medium to large projects together.
What Lunate AI Is Selling: Experience Over Code
Daniel: So this is an advisory business. You’re not sitting down writing Python libraries. What exactly are you selling?
Nadine: Honestly, it’s experience. And here’s something I’m hearing more often: clients can now vibe-code a solution themselves, but then they come back and say, “Can you double-check this? I really don’t know if it’s right.” What we’re selling is accumulated experience that accelerates the pathway. If you’re a government agency incorporating AI, there’s real accountability. If you’re a startup, you’re betting your future on it. If it’s a project, it has to meet a certain quality bar. We bring all of that together — across GOAI, AI assistants, and now agentic orchestration. Primarily, we’re an advisory business. And you have to trust your advisor.
Moving Up the Stack: From Coder to Expert Advisor
Daniel: There are GIS professionals listening right now wondering whether to double down on technical skills — Python, JavaScript — or put their eggs in a different basket. How does someone move from being the person who pushes buttons to becoming the expert advisor?
Nadine: Take a simple example. Google’s AlphaEarth embeddings are now accessible. A non-profit overseas can almost magically identify fields or solar panels from imagery. But somebody still needs to show that customer what this actually is — demystify it, run a few scripts together, let them play. Then come back and ask: which of your use cases would benefit? How would your workflow change? You’re not just coding anymore — you’re showing the value of the whole ecosystem to people who are still standing on the riverbank.
Nadine: The ease of building prototypes today is remarkable. People with no software development background are building usable things quickly. But there remains a huge gap between proof of concept and production. We recently worked with the UK government. I can demo something impressive in two hours. But can a bank run that same thing every morning, trust the results, and know it’ll still be there five years from now? That’s the hard work — and that’s where the real opportunity is.
Daniel: The one-size-fits-all dashboard era has to be over. You can now build personalized dashboards for anyone — and if they don’t like it, scrap it and rebuild.
Nadine: And it goes even further than dashboards. A French startup called Globe Holder recently released a platform where, with generative AI and Type 2 reasoning, you can run full simulations — world models. Instead of a dashboard showing you the past, you run scenarios in the background and surface the most likely outcome: where to put your next power plant, where to locate a distribution centre, where to buy real estate. The dashboard isn’t dead — it’s being replaced by something that can reason about the future.
Trust, Guardrails, and the Gap Between Demo and Production
Daniel: If I’m a bank and this is critical infrastructure, it needs to work every single time — like a bridge.
Nadine: Exactly. And it takes just one failure for trust to collapse. AI isn’t quite the black box everyone feared, but it’s not the fully deterministic system we’re all used to either. You cannot be in a position where it “sometimes works.” The business is accountable. The organisation is accountable. We haven’t even settled the question of legal liability when AI fails — but that conversation is coming.
Daniel: In autonomous vehicles, human driving performance is the benchmark the AI has to beat before we accept it. Do you see that kind of reference standard emerging in geospatial AI?
Nadine: It completely depends on the use case. At NASA, scientific integrity is non-negotiable. They’re using AI workflows in specific contexts, but with scientists reviewing the outputs. Defense applications operate under an entirely different framework than general mapping or insurance analysis. I see two extremes: startups pushing forward fast because they can afford to experiment, and government agencies moving deliberately because they have to navigate accountability, governance, and workforce transformation. NOAA recently launched its first AI pilot for weather forecasting and appointed a Chief AI Officer to oversee it. Their biggest time investment? Defining exactly what they want the AI to do and for whom — and having it tested by actual weather forecasters before anything goes live.
AI Agents and the Future of Satellite Tasking
Daniel: You keep coming back to agents. Why agents specifically?
Nadine: Because that’s genuinely where this is going. Think of it as an evolution: first we had all the data, then web services, then APIs. Agents are the next abstraction layer — instead of orchestrating APIs, we’ll orchestrate agents. Marketing agents, logistics agents, millions of geospatial agents. And there are startups right now building the operating systems and orchestration layers for all of this.
Nadine: A concrete example: I’m working with a company called Extelis that’s launching a new SAR constellation. What fascinates me is that by the time they launch, the business model may simply be: “I’m not selling you data — I’m selling you an AI agent.” There’s already a company called Novi where you upload your algorithm to a portal, and the satellite executes it. You don’t task the satellite, you don’t interact with the data directly. You say “here’s my trusted algorithm, here’s my trusted satellite — send me what I need.”
Daniel: So the agent decides when to task, what resolution to use, when to hand off to SAR versus optical — it just handles it.
Nadine: They’re already doing this. At the Space Symposium I just came back from, you can see agents becoming increasingly autonomous — detecting an anomaly, requesting better resolution from a sister satellite, making the re-imaging decision without human intervention. Which brings it all back to the same point: the real work is guardrails. Proving the demo works is easy. Can a bank trust those results every morning, every day, for years? That’s the hard part. That’s what we focus on.
The Barriers to Modernization
Daniel: What are the biggest barriers when organisations know they need to modernize but don’t know where to start?
Nadine: It’s always a combination. Things are moving too fast — you can’t wait three weeks for your current workflow when a disaster or conflict is unfolding in real time. Decision-makers hesitate to commit, so they fund experimentation, but the gap between experimentation and operationalization remains massive. Behind that gap you find trust challenges, workforce questions, procurement processes, SLA expectations, and cloud readiness. Is your data in the cloud? Are you using cloud-native formats? Sometimes that’s genuinely where you have to start. And then there’s competition — knowing who else is doing this, how defensible your position is, and whether you’re building the fifth version of something that already exists.
Where the Smart Money Goes: A Marketplace of Agents
Daniel: If I handed you a big bag of money, where would you put it right now?
Nadine: A marketplace of agents. Agents for agriculture, agents for insurance, agents for mining. Build them, orchestrate them. That’s where I’d go.
Daniel: Is “agents” the new buzzword — like “big data” was — or is there something genuinely different here?
Nadine: It’s genuinely different. The evolution goes: all the data, then web services, then APIs, then agents. Instead of orchestrating APIs, you orchestrate intelligent autonomous actors. You can already see it in the tools we use every day — Claude Code, for example, spawns multiple agents working in parallel on a problem. I cannot see how this isn’t inevitable. And when I showed a government agency how something that used to take three weeks could be answered in seconds — the looks in the room said everything. That’s what makes this work exciting.
How to Work with Lunate AI
Daniel: If someone listening wants to work with you, what should they do?
Nadine: LinkedIn is the easiest. Or go to lunateai.com — there’s a contact form. I’d love to talk. It’s genuinely a two-way street: we learn a great deal from every conversation, and we bring all of that accumulated experience — lessons learned, partnerships tested, projects that worked and ones that didn’t — back to whoever we’re working with. The goal, eventually, is a whole body of knowledge: maybe a checklist, maybe an agent.
