Industrial AI
Engineered for Impact
My primary focus: knowledge management and AI search for industrial companies — based on TyrolAI Docs. I also consult on computer vision, predictive maintenance, and process optimization.
Knowledge Management & RAG
Stop wasting hours searching for technical documentation, machine manuals, or maintenance logs. I build private Retrieval-Augmented Generation (RAG) systems that let your employees chat securely with all of your company data.
- Chat directly with PDF manuals, CAD specs, and Excel logs
- 100% data privacy: On-Premise or secure Private Cloud
- Microsoft SSO, AD group sync, and document-level security
- Reduce search and onboarding time by up to 80%
TyrolAI Docs — my own enterprise RAG platform based on IBM's OpenRAG, hardened for Austrian industry. The foundation for every knowledge-management project I deliver. Read more →
Enterprise Knowledge Chat
Predictive Maintenance
Stop performing maintenance too early (wasting parts) or too late (machine breakdown). By analyzing vibration, acoustic, and temperature data through Edge AI, I predict component failures weeks in advance.
- Vibration analysis on CNC spindles to detect bearing wear
- Acoustic anomaly detection in complex gearboxes
- Up to 40% reduction in unplanned downtime
Anomaly Detection Score
Predicted Failure in
14 Days
Process Optimization & RL
Machine parameters are often set by gut feeling. I use Deep Reinforcement Learning and Digital Twins to permanently calculate the mathematical "Golden Batch" for your PLCs.
- Dynamic adjustment of cutting speeds based on material variance
- Reducing cycle times by 10-15% without sacrificing quality
- Energy consumption minimization via intelligent scheduling
Computer Vision & Quality Control
Say goodbye to human error in defect detection. I deploy state-of-the-art neural networks (CNNs, Vision Transformers) combined with industrial high-speed cameras to inspect products at speeds and accuracies no human can match.
- Detection of micro-scratches on metal surfaces (down to 0.1mm)
- Dimensional accuracy checks without manual measuring tools
- Classification of organic or unpredictable material defects
def inspect_part(image_tensor):
# Run optimized ResNet50
prediction = model(image_tensor)
defect_score = prediction['defect_prob']
if defect_score > 0.95:
trigger_pneumatic_ejector()
log_anomaly(defect_score)
return "REJECT"
return "PASS"