Founder Keith McCall on building companies, software, and real-world execution.
I was recently reminded how fast the ground is shifting. For years, geospatial software was treated like a fortress: expensive licenses, specialized desktop tools, complex workflows, and a long path from data capture to something useful. That model is breaking.
The moment we became more vocal about Omniris embracing open source technologies as part of our geospatial data platform strategy, something interesting happened: Hardware vendors started reaching out. Not cautiously. Not skeptically. Enthusiastically.
Most teams don’t have a drone problem. They have a data problem. Tools like Pix4D solved something important: turning imagery into outputs—orthomosaics, point clouds, 3D models. That mattered. But let’s be blunt: Stitching is no longer a category. It’s a feature. And features don’t define platforms.
For years, the industry has treated point clouds as the end state of 3D capture. They’re not. They’re precise and essential — but incomplete when it comes to visual fidelity and AI context. That gap is where Gaussian splats step in. I’m pleased to announce that Gaussian splats are now live on the Omniris Vault—turning point clouds into AI-ready, visually rich datasets and unlocking an entirely new class of geospatial workflows.
When we started building Omniris, we didn’t begin with features. We started with outcomes. The vision was clear: a platform that could enable the collection, storage, and management of data from drones, satellites, and IoT devices—across industries. That idea was sparked during our work with Pollen Systems in AgTech, but we knew it had to extend far beyond a single vertical.
If you’re a drone pilot, you’re starting to hear the same thing everywhere: “Can you capture a Gaussian splat?” The demos look incredible. The expectations are rising. And suddenly, what used to be a standard mapping flight now feels like it’s not enough. But here’s the problem: No one is actually telling you how to capture one.
Every drone workflow still looks the same: Plan → fly → upload → stitch → validate → deliver. Stitching is just a feature—one that’s commoditized and pricing to zero in 2026. Which should tell you something important 👇 Stitching is not where the value is. Farewell, products that simply stitch imagery: you’re now a commoditized feature.
Four years. Roughly $2M of my own capital. A few million more of that of our investors. Alignment with a major platform provider. Evangelism. Advocacy. Support. And not a single lead. At some point, founders have to separate belief from evidence. We just did. We’ve moved our default map provider in Omniris to MapLibre.
What Omniris Changes Omniris brings employee pilots and outsourced flight operations into a single platform. We’ve introduced a new capability: 👉 Drone Service Providers onboard as Clients and we promote them to Organizations 👉 They manage their own pilot fleets inside Omniris 👉 You assign work across employee pilots, DSPs, and the 4,000+ Omniris pilot network All in one place.
A day ago, I said Gaussian splats were overrated. This morning, in our weekly review, our CTO pushed back. “They don’t give you the precise metrics you need… but they’re incredibly powerful for generating digital twins.” That’s the difference between opinion and execution.
"A Gaussian splat on the other hand is not raw data. It has nothing to do with LiDAR points. Each point in a Gaussian splat is a tiny coloured ellipsoid with varying size, varying rotation, and varying opacity. It isn't even captured through hardware. You can't collect Gaussian splats directly in the field." - Dr. Pragra Vaishanav
And you hear this: “What do we actually do with all this data?” That’s the blind spot. The lack of a defined outcome.
So when he suggested applying Gaussian splats to my AgTech company, Pollen Systems, it wasn’t theoretical—it was grounded in real use: Analyze farms to optimize water, fertilizer, pesticides, and biostimulants through a continuous, living view of the field. That stuck.
I started developing software on a computer: the Sinclair ZX81, with one kilobyte of RAM. Today, you can carry 20 terabytes on a device the size of your finger. That’s not linear progress. That’s a complete reset in how we think about scale. And we’re seeing the same shift in geospatial.
We took drone captures of Elephant Mountain and pushed them through Omniris Vault—end-to-end. No external tooling. No stitched-together workflows. No handoffs. Just ingestion → processing → 3D output.
“What is this Omniris thing—and can I actually see it?” We're done explaining. The Vault is open. SO GO USE IT.
We’ve made it easy to collect and ingest drone data. But once the data lands… that’s where most systems break. Not because of the drones. Not because of the data quality. Because the pipeline wasn’t built for what happens next.
There’s a gap in how most drone operations are planned today. Pilots bounce between apps—weather, wind, GNSS—and then try to stitch it together mentally. That fragmentation wastes time and introduces risk.
Built for real industries: construction, telecom, energy, agriculture, solar, property, real estate. Check. Not demos—operations. Now we’re looking for partners to build the insight layer—the modules that turn data into decisions—just like we’ve done with Pollen Systems in agriculture. If you’re building where data meets operations, let’s talk.
Mining: Drones are quietly rewriting how we discover minerals. Not with one sensor—but with layered intelligence.
Crossing the Chasm: Geospatial Intelligence Has Left the Old Market Behind We are crossing a threshold at Omniris. Not iterating. Not improving. Breaking away.
What we inherited was not a polished system or platform. It was a company in disarray. We’ve had to work through delayed pilot payments, outstanding obligations, cost structure problems, and the painful realities of integrating a distressed operation. None of that was easy. Some of it was brutal.
Three months ago, we brought on Craig Eisler as CTO. Since then, the pace of execution has fundamentally changed. We’ve moved beyond concept into a fully operational geospatial platform.
Game on. Seven to ten years ago, ag tech ran this exact experiment. And still—no breakout outcome.
Structured geospatial pipelines reveal inefficiencies, artifacts, and hidden value across drone and agricultural data systems.
Publishing, structure, and usability define whether geospatial systems actually create value.
AI and geospatial systems are converging into real-world operational platforms.
Structured, governed geospatial storage is the foundation for scalable intelligence.