Industrial AI Podcast Time Series Data Quality
May 21, 2025
Join Thomas Dhollander, Co-founder and CPO of TimeSeer.AI, as he shares insights from his engineering and applied machine learning background. He discusses the critical role of trusted IIoT data for proactive operations and why data quality often outstrips model complexity. Thomas highlights common pitfalls like sensor failures and metadata issues that disrupt analytics. He also explains how TimeSeer empowers data stewards, enabling effective data management and responsive workflows across organizations. AI-driven solutions and low-code validation processes are game-changers for operational integrity.
AI Snips
Chapters
Transcript
Episode notes
Models Fail Without Trusted Data
- Machine learning models alone won't scale if the underlying data is poor.
- Thomas Dhollander says better data management becomes the key bottleneck as projects move from pilots to production.
Real Sensor Failures On The Floor
- Sensors often behave imperfectly: stale readings, level shifts, drift and spikes when devices reconnect.
- Thomas Dhollander recounts real-world examples like replaced sensors causing unit mismatches and meters spiking after offline periods.
Metadata Breaks Reuse And Discovery
- Poor metadata (units, limits) makes sensor data undiscoverable outside immediate operations.
- Thomas Dhollander warns that lifting data into cloud or central models exposes these metadata weaknesses dramatically.