cv

PhD — High Energy Physics

• Fields: Physics · Statistics · Machine Learning · Data Science
• Doctoral research with ATLAS at CERN on peta-byte scale data analysis and Monte Carlo physics modelling

MSci — Physics

• Fields: Physics · Mathematics

BSc — Physics

• Fields: Physics · Mathematics

Principal Investigator — Machine Learning Lead

• ML lead of NEEDLE (NEural-basEd Diffusion Likelihood Estimations): diffusion/flow-matching neural simulation-based inference toolkit for distributed HPC, aimed at cross-sector Helmholtz adoption (release target ~2027)
• Authored and obtained a €0.5M Helmholtz AI grant to fund NEEDLE software development
• Built Luigi-based DAG orchestration for ML training, evaluation, experimentation, and versioning on HPC/HTC (≥50,000 cores)
• Developed complex-valued flow-matching models for signed/quasiprobabilistic paths, and neural density-ratio evaluation ~10× faster than conventional ODE/SDE solvers
• Manage an ~10-person international team of data scientists, ML engineers, and scientists across institutes

Lead AI Engineer

• Lead AI engineer for automotive code–requirements traceability (Tier-1 / OEM)
• Embedded safety-critical documents into Qwen/Ollama LLMs as hypernetworks with RAG, improving safety-critical query accuracy (~80%), reducing token usage, and cutting response time (>90%)
• Designed LangChain/LangGraph agentic system translating natural language to Cypher for ubTrace (Kuzu, Neo4j, Memgraph)
• Built as-Code MCP architecture for AI agents, reducing LLM token usage ~70% and latency ~40%
• Built scalable infra for parallel production/development AI pipelines (ubCode / ubTrace / ubc)
• Maintainer of the COVESA Generative AI working group (MCP tooling and A2A patterns for automotive)

Research Fellow — Machine Learning Researcher

• Generative AI (diffusion/flow matching), computer vision, and probabilistic ML for Monte Carlo physics simulations
• Diffusion/flow models to emulate particle interactions with 2D/3D silicon sensors (~1M sensors) for LHC detectors; ONNX deployment path toward the WLCG (~1.5M cores)
• Built Open3D/ML point-cloud visualisation combining 3D meshes and point clouds for multi-modal diffusion
• Funded international collaboration with the Croatian Science Foundation via EU Mobility and Resilience Plan grants (~€16k)
• Quasi-probabilistic ML classifiers for quantum physics (and related signed-weight settings); HiDA grant (~€20k) with American Family Insurance Data Science Institute, Wisconsin–Madison
• System architecture engineer for the German National Analysis Facility (NAF) HPC infrastructure planning

Postdoctoral Research Associate — HEP & Machine Learning

• Peta-byte scale analyses and computational optimisation for synthetic data generation
• Project leader of a 20–40 person Monte Carlo / synthetic-data team on the WLCG
• Deployed CARL-torch to reduce exa-byte-scale synthetic-data disk usage by up to ~50%
• Built C++/Python SPEC-2006 benchmarking for ATLAS simulation; adaptive biasing cut Monte Carlo generation time by ~50%
• Developed GNN embedding models for detector image feature clustering

Postdoctoral Researcher — HEP & Machine Learning

• Co-author of open-source CARL-torch (PyTorch likelihood-ratio estimator, CPU & GPU)
• Co-discoverer of key Higgs decay mode: Phys. Lett. B 786 (2018) 59 (observation of H→bb̅ and VH)
• BDT likelihood-ratio importance sampling increasing Monte Carlo statistical precision by >2×
• Core developer of CxAOD ETL analysis software used by ~300–500 personnel across CERN (still in use)
• Project leader of a 5–10 person Monte Carlo simulation team; supervision/teaching of PhD/MSc/BSc students

PhD Research Associate — Data Science & Physics

• Core developer of multi-language (C++/Python/Bash) peta-byte ETL analysis software used by 300–500 scientists on the WLCG
• Monte Carlo integrator for heavy-flavour processes; BDT classifiers for hypothesis testing on LHC data

Research Associate — Medical Physics / Computer Vision

• C/C++ simulation of single-photon counting with electron-bombarded CCDs
• Centroiding algorithms for sub-pixel accuracy (~25% improvement in time/position resolution) for fluorescence lifetime imaging

Calibrated Density Ratio Estimation for Diffusion Models (PMF-GOOUT-2026-03)

• Host and PI of a 6-month project on fast neural likelihood evaluation for diffusion models in LHC physics

NEEDLE — Neural-based Diffusion Likelihood-free Estimations

• Author and PI of a 3-year collaboration (ATLAS DESY & CMS KIT) on neural likelihood-free inference for HEP
• €0.5M Helmholtz AI funding

Neural Likelihood-free Inference in Peta-byte Data Analyses (MOBDOK-2023)

• Collaboration with University of Zagreb on neural likelihood-free inference for ATLAS & ALICE

Interpreting Negative Probabilities in Supervised Learning

• Author, host, and PI of a 3-month HiDA project on quasiprobabilistic likelihood-ratio estimation

• Institute of Physics Machine Learning: Science and Technology Outstanding Reviewer Award

• SET Award Finalist — Institute of Physics (1 of 3 undergraduate physics projects nationally)
• James Clerk Maxwell Prize — King's College London (excellence in 4th-year thesis)
• Layton Science Research Award — King's College London
• Wheatstone Prize — King's College London (highest 4th-year grade)

• Nelkon Prize — King's College London (highest 3rd-year / BSc-equivalent grade)
• Leica Microsystems Prize — King's College London (highest 2nd-year laboratory Physics)

ATLAS National Analysis Facility (NAF) Representative — Germany

Leader, Machine Learning Assisted Monte Carlo Uncertainties forum — ATLAS

Higgs Machine Learning Expert/Liaison & Paper Reviewer — ATLAS

ATLAS Physics Modelling (Weak Vector Boson) & Monte Carlo Simulation Convener

ATLAS Hbb Physics Modelling & Monte Carlo Simulation Expert/Contact

ATLAS VH(bb) Analysis leadership member

Helmholtz AI Conference 2026

• Organisation Committee Member — haicon.cc

Helmholtz AI Conference 2025

• Abstract Reviewer — haicon.cc

2nd ATLAS Higgs Workshop 2023

• Machine Learning Program Organiser — Indico

Helmholtz AI Conference 2023

• Abstract Reviewer; S-03b Applied AI Session Convener

ATLAS H→bb Workshop

• Monte Carlo Physics Simulation Program Organiser

NEEDLE — Columnar Workflow Orchestrator for Large-Scale Neural SBI in HEP

Neural Quasiprobabilistic Likelihood Ratio Estimation with Negatively Weighted Data

Quasi-probabilistic LR estimation; DeGeSim diffusion density mappers

DeGeSim — Conditional DDPMs as Multi-Dimensional Density Mappers

ATLAS Higgs Group Machine Learning Overview

Novel experimental techniques for the Higgs sector

Higgs measurements in 3rd-generation decay channels (ATLAS & CMS)

Terascale Statistics School — Machine Learning lecture series

DESY Summer School — Machine Learning lecture series

Yearly ATLAS Germany ML lecture series (~100 BSc/MSc/PhD/postdocs)

• Interactive tutorials on machine learning for HEP data analyses

Laboratory teaching (FP1 & FP2) — Universität Freiburg

• Designed and built 3 experiments; entrance exams, report grading, and scientific writing feedback for BSc/MSc students