Cyprien Hoelzl

Data scientist & infrastructure researcher

cyprienhoelzl[at]gmail.com LinkedIn Zürich
Core experience Research Selected projects Self-hosted services
I build models that predict failures before they happen. Trained as a civil engineer, spent my PhD studying how bridges and railways age, now applying similar approaches to risk analysis and operational forecasting. Former co-founder at Irmos, where we built anomaly detection systems for distributed infrastructure.
6+ yrs
Data science
Co-founder
Irmos AG
ETH PhD
Infrastructure

Core experience

Predictive modeling

Time-series forecasting and anomaly detection on high-frequency sensor data. Built early-warning systems for infrastructure and operational failures.

Risk analysis

Monte Carlo simulation, Bayesian inference, stochastic processes. From railway maintenance scheduling to investment risk assessment.

Team leadership

Led cross-functional projects as Head of Innovation. Managed stakeholder reporting, contract negotiations, and delivered systems under tight deadlines.

Production ML

Deployed scalable pipelines on GCP (Python, TensorFlow, BigQuery). Built dashboards translating technical metrics into business decisions.

Research

Railway weld monitoring

Fusing expert knowledge with sensor data for condition assessment. Published in Sensors, 2023.

Read paper

Reinforcement learning for maintenance

Comparing value-based and policy-based RL for infrastructure planning. Published in Structural Health Monitoring, 2023.

Read paper

Selected projects

Real-time anomaly detection at scale

Built distributed monitoring system processing sensor data across multiple sites. Deployed ML models detecting equipment failures 24-48 hours before occurrence. Reduced downtime by enabling proactive maintenance.

Technical details

Risk dashboards for data informed decision making

Created analytics platform combining operational metrics with data driven indicators. Monte Carlo simulations provided confidence intervals for ressource allocation.

Case study

Railway infrastructure deterioration models

PhD research using Bayesian networks to fuse expert knowledge with inspection data. Developed algorithms estimating remaining service life of railway welds. Collaborated with Swiss Federal Railways.

Read paper

Smart garment sizing recommendation

ETH Entrepreneurship Award 2020. Use camera, accelerometer and gyroscope data to estimate body measurements. Reduced fashion e-commerce return rates in pilot study.

Wind turbine fatigue estimation

Applied unsupervised learning to identify operational modes from turbine sensor data. Estimated lifetime consumption through fatigue damage accumulation models.

Home automation power optimization

Adøn Power Down project. Developed user detection system to automatically cut standby power consumption. Combined motion sensors with usage pattern learning.