Kajal Negi

Evaluated MLLM mathematical reasoning across four dimensions and curated two original datasets to support benchmarking.

Extracted the part of the text associated with a label using label information and refined the results for better match and complete phrase.

Automated identification and categorization of reported shortcomings and created new defect categories based on data analysis.

Took average normalized ratio of lap time historic data across seasons and improved predictions of number of laps considering real-time incoming data.

Performed trending theme extraction analysis from textual documents and built sentiment analysis classifiers and summarization for text documents.

Captured fluctuation and drop rate in sensor generated data to predict unexpected breakdown of gas turbines. Ensembled 100 logistic models to capture anomalous data sets, achieving sensitivity of 54% and specificity of 84%.

Derived variables reflecting frequency, quantity, quality, and amount of historic purchases to predict coupon purchase. Fitted LSTM to achieve a ROC of 84% on the test dataset.

Constructed constraints based on business requirements for scheduling and allocated weights corresponding to store and year date pairs using TSP and previous inventory dates for constraint programming.

Ph.D. in AI at the University of Pisa and CNR-ISTI, focusing on evaluating MLLM mathematical reasoning and curating datasets for benchmarking.

Master's degree in Mathematics from the Indian Institute of Technology, Kanpur.

Bachelor's degree in Mathematics from the University of Delhi, Daulat Ram College.