Dr. Baishali Dutta

Developing a solution to extract the most important keywords/phrases that provides a quick and easy overview of customers’ feedbacks for a certain product or product category from large-scale reviews instead of going through the reviews one-by-one. It would help companies to analyse their product/brand status while analysing where they excel or lack in comparison with their competitors. Involved technologies are KeyBERT model using BERT-embeddings, spaCy, SQL, AWS and Docker.

Searched for Dark Matter in hundreds of petabytes of data collected at the biggest CERN detector. Used several machine learning algorithms primarily boosted decision tree and neural networks for feature extraction and hypothesis testing to interpret the data. Publication: https://arxiv.org/abs/2011.09308

* Calculated the cross section, the probability of the W+W- process to happen on the enormous ATLAS dataset * Performed minimisation of the negative log-likelihood function on the carefully selected signal of interest while tackling the correlations * Derived actionable insights from massive datasets with very minimal support * The result proved to be a validation of our physics knowledge and an essential input to various theoretical models Publication: https://cds.cern.ch/record/2635975?ln=en

Performed data acquisition and cleaning on the data collected by the Scintillator detector from various sources. Appliedstatistical regression to interpret and improve the detector performance

Doctoral Thesis at CERN: Performed a precision measurement using the massive CERN dataset. Applied advanced statistical methods such as likelihood profiling to carefully select the signal of interest and derived actionable insights with very minimal support.

Master's Thesis on Fermilab, USA Project: Studied the expected performance and potential of a future Fermilab experiment using simulations. Used data science techniques primarily chi-square ($\chi^2$) minimization to determine the sensitivity of the experiment.

Honours in Physics