Valentina Vecchio, PhD Researcher & Data Scientist
Profile

I am a scientific researcher with 10 years of experience in data analysis. My expertise lies in developing statistical methods to extract insights from large datasets. I am proficient in the latest analysis techniques, including classical Machine Learning and more modern Deep Learning algorithms.

I am a strong communicator, able to convey complex concepts both to peer scientists and to laymen.

I am a highly influential player in the ATLAS collaboration, a international scientific collaboration at CERN which involves roughly 4,000 members. I am recognised for my intellectual contributions to high-profile analyses on the Higgs Boson and for covering leadership roles of multiple research teams totaling around 200 members.

Skills
Data Analysis & Machine Learning
Git | CI/CD | MkDocs
Numpy | Pandas | Polars
Machine Learning & Deep Learning
Python | C++ | Bash | LaTeX
Scikit-learn | PyTorch | XGBoost
Professional Experience
11/2019 – presentManchester, UK
University of Manchester, Research Associate

I work with the ATLAS collaboration, performing research on the interactions between elementary particles.

I have led the implementation of a physics-informed transformer network for particle identification (known as flavour tagging):

  • •achieved 4x performance improvement over old, non-transformer implementation,
  • •network was trained on simulated data of particle collisions,
  • •together, me and my team covered the full development pipeline - from data preprocessing (removing bias using re-sampling and re-weighting), training using PyTorch Lightning, validation/characterisation of network performance and production-grade deployment using ONNX runtime,
  • •training data set (250M) and model size (2.3M) was limited due to resources constraints. We achieved performance goals (and improved network interpret-ability) by enhancing the transformer with physics-informed auxiliary tasks.

    • I contributed to the dissimination of this work by writing:

  • •peer reviewed paper accepted for publication in Nature Communications;
  • •briefing article for outreach aimed to the general public.

    • Developed a data analysis for the study of the Higgs boson properties (extracting signal in a background-dominated environment):

  • •Used multi-class boosted decision trees for the classification of signal vs background
  • •Used binned profiled likelihood fit for both measuring signal production rate and data-driven estimation of background contribution.
    • Valentina Vecchio, PhD
    1 / 2
  • •Implemented bootstrapping method to quantify the impact of statistical fluctuations on systemic uncertainties.

    • I contributed to the dissimination of this work by writing the peer reviewed paper accepted for publication in JHEP.

    07/2018 – 06/2019Geneva, Switzerland
    CERN, Associate Fellow

    This fellowship is usually granted to PostDoctoral Researchers and only exceptionally to PhD students who distinguished themselves for their work. I was awarded this fellowship in the last year of my PhD for my contributions to top‑quark precision measurements and data‑driven calibration of flavour tagging performance within the ATLAS experiment.

    International Leadership

    During my time as an RA I have played various leadership roles of which the most senior has been convening a group of 200 researchers and PhD students across multiple institutions around the world. Our work is used by 90% of the data analyses carried out within the ATLAS scientific collaboration at CERN.

  • •Provide strategic guidance to the group to meet the collaboration's broader goal of driving innovative research and discovery.
  • •Liaise between senior stakeholders and team members to translate strategic priorities into achievable deliverables and define measurable benchmarks to assess analytical effectiveness.
  • •Mentorship of several PhD/MSc Students.
  • •Lead the articulation and communication of compelling narratives to explain our scientific findings both to the wider research community, to funding stakeholders and to the public.
    • Education
    10/2016 – 10/2019Rome, Italy
    PhD in Physics, Università degli Studi Roma Tre
    10/2014 – 09/2016Rome, Italy
    Masters in Physics [110/110 Cum Laude], Università degli Studi Roma Tre
    10/2011 – 09/2014Rome, Italy
    Bachelor in Physics [110/110 Cum Laude], Università degli Studi Roma Tre
    Talks & Workshops

    I regularly present at international conference and workshops of the sector. I took part to the organisation of several international topical workshops on the usage of machine learning for particle physics experiments. A complete list can be found here

    Grants & Awards
    Co-Leader & UK Representative, EU COST-Action: Machine Learning and Quantum Computing for Future Colliders (MLQC4FC)
    2025
    Valentina Vecchio, PhD
    2 / 2