Data Scientist

Technical Skills: Python (incl. scipy, numpy, pandas, scikit-learn, jupyter notebook, streamlit), Mathematica, Linux, Mac, GitHub, C++, LaTex

Education

• Habilitation, Physics

  The University of Graz (June 2021)

• Ph.D., Physics

  The University of Bonn (August 2007)

• M.Sc., Physics

  The University Mumbai (May 2006)

Work Experience

Group leader @ University of Graz, Austria (September 2020 - Present)

• Research group head: Led a team connecting theoretical models with experimental collider data.
• Managed €1M in project funding and supervised 4 researchers (2 PhDs, 1 MSc, 1 BSc).
• Designed and taught 4 graduate-level courses on data modeling and theoretical physics.
• Delivered 20+ invited talks and published 10+ peer-reviewed papers.

Elise-Richter Fellow @ Austrian Academy of Sciences, Vienna, Austria (September 2017 - August 2020)

• Third party funding of approx 390K Euro
• Statistical analysis, data visualisation, quantitative comparison with predictions of theoretical models of dark matter
• Lead developer of software tool SModelS
• Supervision, mentoring of five master students

Postdoctoral Research Fellow @ Austrian Academy of Sciences, Vienna, Austria (September 2017 - August 2020)

• Conducted research on dark matter and neutrinos
• Organised workshops and conferences
• Supervised master students
• Presented research at conferences and seminars
• Participated in international CERN Large Hadron Collider collaboration

Postdoctoral Research Fellow @ CNRS, Grenoble, France (September 2011 - August 2024)

• Conducted research on dark matter
• Organised seminars
• Supervised master students
• Presented research at conferences and seminars
• Lead developer of public data analysis code SMoldeS written in Python

Data Science Projects

Time series analysis - Remaining unit life for NASA dataset

• Documentation and code
• Built a XGBoost and Optuna integrated regression analysis to predict remaining unit life for NASA engines
• Aims:
  • Built consistent time series analysis using sliding window techniques
  • Used correct labeled dataset to preserve the unit structures
  • Tuned hyper-parameters using Optuna and estimated quantiles

Anomaly Detection – LSTM Autoencoder App

• Documentation and code
• Built a Streamlit demo using LSTM autoencoder to detect anomalies in physical system signals
• Aims:
  • To understand and explain role of several hyper-parameters in an LSTM anomaly detection network
  • To construct a first physics inspired neutral network

PCOS diagnosis

• Documentation and code
• Interactive plots for data visualisation
• Aims:
  • To understand correlation analysis
  • To work with binary data
  • To create a first version of PCOS prediction system using random forrest classifier

Taylor Swift music analysis and content recommendation system

Data analysis of Taylor Swift’s music (EDA)

• Documentation and code
• Initial assessment of Taylor Swift spotify music records
• Aims:
  • To undestand the nature of data and perform statistical analysis to gain insights into correlations
  • To create basic visualisations and summaries in form of histograms, scatter plots and pandas summary dataframes

Data analysis of Taylor Swift’s music (Content based recommendation system)

• Documentation and code
• Interactive plots for data visualisation and clustering analysis for song suggestions
• Aims:
  • To understand the behaviour of widgets and correspondingly use it to create interactive plots
  • To perform elementary clustering exercises via KMeans algorithm in order to predict songs the user might want to listen to given their choice of song.

Publications

Search for long-lived particles decaying to a pair of muons in proton-proton collisions at \(\sqrt{s} = 13\) TeV

Publication

Performed numerical simulations and created visualisations using Python for search optimisation. This led to more refined definitions of search parameters and impacted fundamental search and data collection design.

Theory, phenomenology, and experimental avenues for dark showers: a Snowmass 2021 report

Publication

Co-ordinated a team of 50+ scientists to collect, organise and analyse the field of specific dark matter models. Created and edited final report design. Contributed to the report by bringing in fundamental understanding of these dark matter models and provided new direction to the field.

Constraining new physics with SModelS version 2

Publication

Designed a new and improved version of public code SMoldeS written in Python. The code accounts for over hundres new physics searches and accelerates the process of understanding the nature of dark matter.