Research

Ongoing PhD analysis of Higgs boson decays to gluons (H → gg), reconstructed as a dijet system, in association with a Z boson. The work uses CMS data and focuses on maximising signal–background separation with Boosted Decision Trees.

Analysis highlights

• BDT training Signal–background classification with XGBoost and TMVA.
• Jet selection Quark–gluon likelihood (QGL) discriminants to identify gluon-like jets as Higgs candidates.
• Analysis framework RDataFrame-based workflow for fast, scalable event processing.

• Python
• C++
• XGBoost
• TMVA
• RDataFrame
• CMS

Developed a complete analysis pipeline for the Higgs boson decaying into two W bosons in the opposite-sign, different-flavour (electron–muon) final state via the ggH production channel. The project uses CMS 2016 Ultra-Legacy Open Data to probe Standard Model physics while demonstrating the scope of CERN Open Data for education and research.

Analysis highlights

• Analysis pipeline End-to-end reconstruction of H → WW → e±μ∓ (ggH) using CMS Open Data, from trigger selection to final-state object definition.
• Object reconstruction Lepton identification, isolation criteria, and missing transverse energy (MET) reconstruction for neutrino inference.
• Signal extraction Kinematic and topological selections applied to isolate signal from dominant backgrounds (top and Drell-Yan).
• Statistical analysis Signal significance estimation with evaluation of statistical uncertainties in the low signal-to-background regime.
• Reproducible workflow Documented and modular analysis chain for HSF-India, emphasizing transparency, corrections, and known limitations.

Technical implementation

• Core stack Uproot (I/O), Awkward Array (jagged structures), Vector (4-vector arithmetic), Hist (yield accumulation).
• Distributed computing Dask for scalable parallel processing of millions of NanoAOD events.
• Statistical inference CMS Combine for simultaneous profile likelihood fits.
• Scale factors HLT trigger, lepton ID, and isolation corrections applied to simulation weights.

• Python
• uproot
• awkward-array
• Dask
• CMS
• NanoAOD
• HSF-India

Master's thesis centred on the Drell-Yan process at the LHC, working with Run 2 Monte Carlo data from the CMS experiment to study Standard Model processes. The analysis focused on the Z → e⁺e⁻ channel, isolating a clean Z boson signal from dominant backgrounds.

Analysis highlights

• Signal extraction Precision reconstruction of the Z → e⁺e⁻ peak (60–120 GeV) using CMS Open Data with MVA-based electron identification (Fall17 V2).
• Event selection Optimised kinematic cuts and lepton isolation criteria to enhance signal purity and data quality.
• Background suppression Projected MET algorithm exploiting lepton–MET angular correlations to suppress tt̄ and diboson backgrounds.
• Analysis framework C++/ROOT-based pipeline using TChain and TLorentzVector with custom data slimming for efficient multi-file processing.
• Validation Shape-based comparison with six MC backgrounds (tt̄, WW, WZ, ZZ) via normalized data–simulation overlays.

Technical implementation

• C++-based ROOT scripts for event filtering, selection, and histogram production.

• C++
• ROOT
• CMS Run 2
• Monte Carlo
• Drell-Yan