Project Overview
| Role: Developer & Analyst | Duration: 2021-Present | Data Volume: TB-scale per campaign |
The Challenge
Process and analyze TB-scale datasets from particle beam experiments with quick turnaround to guide ongoing data-taking decisions. Manual analysis took weeks; beam campaigns run on tight schedules with limited access.
Solution
Developed the LGADUtils C++/ROOT analysis framework with modular architecture:
- WaveformAnalysis: Oscilloscope parsing, baseline subtraction, signal extraction
- TimingExtraction: Configurable constant-fraction discriminator (CFD) algorithm with per-pixel timewalk correction
- TrackMatching: Spatial and temporal correlation between telescope tracks and device-under-test hits
- CalibrationTools: VCO frequency calibration, drift correction, charge calibration via test-pulse injection
- GRID processing: 185,000+ HTCondor jobs for timing characterization across campaigns
- Real-time monitoring: Data quality checks at ~50 MB/s TCP throughput, near-100% live-time during beam spills
Technical Stack
Python ROOT Pandas NumPy Git Bash HTCondor GRID Computing
Results
- Reduced analysis turnaround from weeks to days across 12+ campaigns
- 185,849 grid jobs processed for the Timepix4 timing characterization alone (3rd highest usage at Nikhef)
- Framework adopted by collaborating institutions within the EP-R&D programme
- Enabled same-day feedback during beam time, directly informing detector configuration decisions
Industry Relevance
Skills directly applicable to:
- Data Engineering: Building robust data pipelines
- DevOps: Automation and workflow orchestration
- Scientific Computing: Statistical analysis and visualization
- Distributed Systems: GRID/cluster computing experience