§ Work · vii. · Alynix

Bridge inspection at the speed of a drone.

Decker replaces a slow, lane-closure-heavy hammer-sounding inspection with a thermal flight, an automated processing pipeline, and a defect map a structural engineer can sign.

Client: Alynix
Year: Multi-year, ongoing
Role: Architect → CTO
Outcome: In production with state DOTs

The premise

The way the United States inspects its bridges in 2026 looks a lot like the way it inspected them in 1976. An engineer, a hammer, a chain, lane closures, traffic delays, and a clipboard. The data product at the end of all of it — a deck map showing where the concrete is delaminating — is essentially the same data product a calibrated thermal camera in the air can produce, faster, safer, and at a cost that finally makes routine inspection economic at scale.

Alynix's bet was that the bottleneck wasn't the imaging. It was the platform around it — capture planning, processing, defect detection, and reporting that fit into how state DOTs already work. That's what we built.

The pipeline

01 · Plan

Flight planning

Per-bridge mission planning that accounts for span geometry, sun angle, thermal contrast windows, and FAA constraints — so a pilot shows up with a flight that's actually going to produce usable data.

02 · Capture

Thermal flight

A thermal-equipped drone flies the deck under conditions where subsurface delamination shows itself thermally. No lane closures, no jackhammers, no traffic control plan.

03 · Process

Register & detect

Frames are stitched, geo-registered, and orthorectified into a single deck mosaic; a detection model identifies thermal anomalies consistent with delamination and bounds them as defects.

04 · Report

DOT-ready deliverable

A structured deck-condition report that fits a state DOT's existing inspection workflow — not a dump of raw imagery, a thing an engineer can review and sign.

Regional macroplan: a state-scale view of bridges planned and collected across a corridor — Fig. 01 — regional macroplan. Hundreds of bridges across a state corridor, planned and collected as a single fleet operation.

Flight planning over a cluster of bridges, with capture footprints and coverage overlays — Fig. 02 — flight planning, up close. Mission geometry, capture footprints, and live coverage preview over a cluster of bridges.

A thermal ortho-mosaic of a bridge deck with detected anomalies highlighted — Fig. 03 — the thermal mosaic. Subsurface delamination shows itself as bright clusters against the deck's thermal signature.

Deck assessment report with detected delamination overlaid on aerial RGB imagery, plus deck and data-quality metrics — Fig. 04 — the deliverable. Detected delamination registered onto an aerial RGB ortho, with deck assessment, collection grade, and data-quality metrics built in.

My role

I started on Decker as the architect of the platform through Lofty Labs — mapping out the system, the data flow, the team, and the build. Over the course of the engagement, my role with Alynix expanded: I transitioned into a CTO seat to help drive go-to-market alongside the founders, taking on commercial conversations, customer engagements, and the technical credibility that selling into state DOTs requires.

The engagement remains ongoing. The core platform is a Django + PostGIS application with a Vue.js operator UI; the heavy work happens in a Python image-processing stack — OpenCV, GDAL, NumPy, scikit-image — with PyTorch-based detection models on top, all wrapped in a workflow that a non-developer field engineer can drive.

What changed

Decker is in production use with state DOT clients today. The category — remote, sensor-driven infrastructure inspection — is moving from interesting research to procured service, and Alynix is one of the small number of companies actually delivering it on real bridges.

Building a sensor-to-deliverable product?

The path from raw sensor capture to a deliverable an industry buyer will sign for is its own discipline. If that's the path you're on, get in touch.

Start a conversation →