About
Built by people who ran the line
Shelfmark was founded in Pittsburgh in 2023 by a plant quality lead who spent years watching defects complete their journey to scrap because the monitoring tools couldn't see them forming.
Company facts
- Founded
- 2023, Pittsburgh PA
- Funding
- $1.5M angel round, January 2026
- Stage
- Early-access pilots active
- Contact
- [email protected]
The problem was always the 90 seconds
When Pat O'Donnell was running quality for a precision wire manufacturer in the Pittsburgh area, the hardest part of the job wasn't finding defects. It was watching them happen, knowing they had probably started forming 90 seconds earlier somewhere upstream, and not having a system that could tell him where or why in time to do anything about it.
The process historians had all the data. The sensors were generating the signal. But nothing was correlating the channels in real time, nothing was comparing live patterns against production baselines, and nothing was telling the floor what to adjust before the spool was committed.
Shelfmark was built to close that gap. Not by replacing the instrumentation that plant engineers have spent years configuring and trusting, but by adding a fusion and scoring layer that makes the signal in that instrumentation visible before the defect is irreversible.
The product is designed for people who have already spent time on a line and understand that false alarms are as damaging as missed detections. We calibrate to your baseline, not to a generic threshold, and we review the alert history with you every month to keep the signal-to-noise ratio where your team needs it.
Team
The people behind the platform
Pat O'Donnell
CEO and Co-Founder
Pat's background is in plant quality engineering on continuous-flow manufacturing lines, where he developed an understanding of how defects actually form, travel, and accumulate before they surface at inspection. He shapes how Shelfmark's alert thresholds, review cadences, and root-cause attribution logic are designed to work on the floor, not just in a test environment.
Marta Kowalski
Head of AI/ML
Marta's background is in time-series anomaly detection and sensor fusion for industrial process environments. She has built monitoring models for continuous manufacturing and process-control applications where false-positive rate directly determines whether operators trust the system. At Shelfmark, she leads model architecture, calibration methodology, and the detection engine that drives real-time defect scoring.
Darnell Foster
Lead Systems Engineer
Darnell's background is in building edge data-collection systems and low-latency data pipelines for manufacturing environments, with hands-on experience connecting OPC-UA, Modbus, and historian layers to plant control networks. At Shelfmark, he owns the ingestion layer, protocol adapters, and deployment architecture that allows connection to existing line instrumentation without hardware changes in most configurations.
Backing
Angel-backed and operationally focused
Shelfmark raised a $1.5M angel round in January 2026 from investors with manufacturing operations backgrounds. We are not chasing enterprise software scale at the expense of the product's usefulness on the floor.