Background
Mipmap3D serves surveyors, architects, and construction teams who depend on photogrammetry—turning thousands of drone images into accurate, georeferenced 3D models. But the dominant tools in this space, Pix4Dmapper and DJI Terra, are built around expensive perpetual licenses and desktop hardware, not the cloud-native economics modern teams need.
As Mipmap3D's client base grew and enterprise contracts came into view, the team knew their path forward couldn't rely on per-seat licenses and on-premise machines. They needed their own platform: one that could scale elastically, process projects faster than the incumbents, and give them full control over the product roadmap. They brought us in to build it.
The Challenges
- •Pix4Dmapper and DJI Terra licensing costs became prohibitive as the team scaled to more projects and users
- •Desktop-bound workflows made it impossible to handle concurrent large-scale jobs without dedicated high-spec hardware per user
- •Processing turnaround times were too slow for competitive project delivery, especially for large drone survey datasets
- •No path to white-labeling or embedding the workflow into their own client-facing product
- •Enterprise prospects expected a modern SaaS experience—not a desktop tool wrapped in a manual process
“We weren't in a crisis, but it always felt like we were one big project away from hitting our limits. We needed a platform we owned.”
Our Approach
We designed the platform in three layers: a precise aerial triangulation pipeline, elastic cloud infrastructure, and a production-ready product surface.
Phase 01: Aerial Triangulation & Camera Calibration Pipeline
The foundation of accurate 3D reconstruction is correct camera and GPS alignment. We built a pipeline that ingests raw drone imagery, performs automated camera calibration using intrinsic and extrinsic parameter estimation, and fuses GPS telemetry to anchor the model in real-world coordinates. The pipeline is optimized for throughput—parallelizing image matching across nodes so large datasets don't become bottlenecks.
- •Automated intrinsic/extrinsic camera calibration per flight session
- •GPS fusion with EXIF and telemetry data for geospatial anchoring
- •Parallel image-matching pipeline across distributed compute nodes
Phase 02: Elastic Cloud Infrastructure on AWS
We moved all compute to AWS, containerized with Docker and orchestrated via ECS. Jobs are dispatched from a queue, and worker nodes scale up automatically to match load—then scale back to zero when idle. This means a surveying firm can submit 10 large jobs simultaneously without pre-provisioning hardware, and pay only for the GPU-hours actually consumed.
- •AWS ECS + Docker for containerized, stateless processing workers
- •Job queue architecture (SQS) for concurrent multi-project processing
- •Auto-scaling worker pools—zero idle cost between projects
Phase 03: Enterprise-Grade Product Surface
We built a React/Next.js frontend backed by a Python + Node.js API layer. Users upload drone imagery, monitor processing in real-time, and download deliverables—orthomosaics, point clouds, DEMs—without touching a command line. The architecture supports multi-tenant project isolation, role-based access, and white-label deployment for Mipmap3D's own enterprise clients.
- •React + Next.js dashboard with real-time job status and progress
- •Python processing backend (OpenSfM-based pipeline) + Node.js API gateway
- •Multi-tenant architecture with role-based access and white-label support
The Results
- •60–80% cost reduction per project vs. Pix4Dmapper and DJI Terra licensing
- •Processing time reduced by processing jobs in parallel across cloud workers
- •Zero upfront hardware investment—elastic infrastructure scales to any project size
- •Enterprise-ready: multi-tenant, white-labelable, deployable for client teams
- •Full ownership of the product roadmap—no dependency on third-party licensing
“For the first time, we're not just surviving on legacy tools—we're scaling on our own platform. That's a game-changer for how we pitch to enterprise clients.”
Final Takeaway
The right architecture doesn't just cut costs—it opens new business models. By moving 3D reconstruction to the cloud and building a product layer on top, Mipmap3D can now compete with incumbents on price, speed, and UX simultaneously. They don't just use software anymore—they sell it.
Technologies We Use
Modern, proven technologies to build robust applications
React
Next.js
Node.js
Python
AWS ECS
AWS SQS
Docker
TypeScript
OpenSfM
Qt
See More Work
Visitors
Activity
easyvisas.ai – Fully AI-Powered Immigration Processing
End-to-end AI automation for visa applications, from document intake to decision support.
I've had a headache for 2 days. When should I see a doctor?
Monitor your symptoms. Seek care if pain worsens, fever appears, or vision changes. I'll track this in your profile.
What vitals should I log daily for heart health?
Log BP, heart rate, and activity. Here's a simple tracking routine.
AI Nurse • Symptom monitoring • Not medical advice
AI Nurse – Predictive Symptom Monitoring System
Real-time symptom tracking and predictive alerts for proactive patient care.
Engineering Partners for Products That Ship
From applied AI and resilient cloud platforms to full-stack delivery—we help you scope with clarity, build with rigor, and release with confidence.