Eugene Pik
-
Email
eugene.pik@essex.ac.uk -
Location
Colchester Campus
Profile
Biography
Eugene Pik is a PhD research student in Data Science within the Department of Mathematics, Statistics and Actuarial Science at the University of Essex. His research applies statistical modelling and large-scale data engineering to the integrity of decentralised aviation surveillance networks, working with datasets of hundreds of billions of records. His methodological interests include count regression for overdispersed and heavy-tailed data, exposure-offset models for rate normalisation, spatial and spatiotemporal analysis, anomaly detection, and the treatment of survivorship bias and pseudoreplication in observational data pipelines. He is the lead author of the Global Envelope Thresholding (GET) framework, which established the first global, traffic-normalised baseline of ADS-B kinematic consistency from 246.7 billion messages using negative binomial regression across 260 Flight Information Regions, and defined "Ghost Records" as a new class of surveillance anomaly. The resulting GET-25 benchmark dataset is openly available with a Python API. His work has appeared in the Journal of Aerospace Information Systems (10.2514/1.I011527), the Journal of Transportation Security (10.1007/s12198-025-00323-w), and AIAA conference proceedings. Eugene holds an M.Sc. in Aviation and Aerospace Sustainability / Aviation Cybersecurity from Embry-Riddle Aeronautical University, with earlier training in robotics and software engineering. He is the founder of Mevocopter Aerospace and previously led requirements for a hybrid-electric eVTOL aircraft programme for emergency medical services.
Qualifications
-
M.Sc. in Aviation and Aerospace Sustainability / Aviation Cybersecurity Embry-Riddle Aeronautical University (2024)
Research and professional activities
Thesis
Kinematic Integrity and Security in Automated Aviation Systems
A quantitative pipeline for anomaly detection and inference on global open ADS-B data. NLP classification maps unstructured safety text to standardized taxonomies; operational anomaly metrics are validated against labelled ground truth; and traffic-normalized rates are modelled via logistic and negative-binomial GLMs with exposure offsets, handling severe class imbalance and overdispersion. Contributions: reproducible metric definitions, cross-source record linkage, and open artefacts.
Supervisor: Igor Rodionov