Aastha Tandon

I am Graduate student at California State University East Bay, pursuing an MS in Business Analytics with a specialization in Machine Learning and Data Science.

My journey so far has led me to explore cutting-edge applications. As a Research Assistant in the Summer of 2023, I dived into advanced NLP techniques, leveraging OpenAI's LLM GPT-3.5 Turbo for structuring data and harnessing BERT and LLM for modeling. I am currently working on the precise fine-tuning of LLM for a specific task.

My educational foundation includes a Bachelor’s in Computer Science and a Master’s in Operation Research from the University of Delhi, India. With 2 years in a healthcare startup CHEORS, I contributed analytics solutions driving efficiency in drug development. Later, during my time at UnitedHealth Group, I focused on ML models to prevent customer churn and enhance breast cancer detection and various other disease areas.

Email  /  Github  /  LinkedIn

My research interests are in computer vision, machine learning and natural language processing, data analytics.

This project deploys Convolutional Neural Networks (CNNs) to detect Invasive Ductal Carcinoma (IDC) in breast histopathology images, addressing class imbalance, optimizing models, and evaluating performance with F1 score, precision, and recall, utilizing cosine learning rate decay for efficient convergence.

Employed LDA topic modeling to analyze customer reviews, identifying key themes in positive and negative feedback, exploring monthly and seasonal variations. The aim is to enhance visitor satisfaction and make informed, data-driven decisions based on customer insights.

Addressed Toxic Comment Classification, employing both traditional ML and LSTM. Explored binary, chained binary, and multiclass logistic regression for label prediction. Additionally, utilized an LSTM model with Glove embeddings for comparison.

Efficiently extracted and processed USAF weather station IDs and average wind direction, range of ceiling heights, and average visibility distances from the NCDC dataset using Mapreduce, Hadoop,Spark,Hive,Pig with AWS and University Hadoop to identify yearly trends.

Addressed telecom operators' customer churn issue by developing a precise predictive model. It includes data preprocessing, demographic and usage pattern analysis, diverse machine learning model usage, and provides a comprehensive report for effective churn management.

Created a robust CNN-based model for classifying 156 orchid species from a dataset of 7156 images, improving efficiency in botany and horticulture. Two approaches are employed: creating CNN models from scratch and fine-tuning pre-trained models. Additionally, a multi-task model predicts both species and flower counts.

Implemented a sequence-to-sequence (seq2seq) langauge translation model for machine translation, aiming to effectively translate text from english to french.

Conducted a comprehensive analysis of Bay Area real estate data gathered through web scraping, revealing key trends impacted by events such as the pandemic and economic changes, empowering stakeholders to make well-informed decisions.

Studied customer satisfaction and their perception of a brand’s apparel. Analyzed various customer segments with demography and the needs of each segment, their satisfaction and assessed their attractiveness towards the brand.

Created a Database Management System for a hotel chain to monitor daily operations, including booking, reservations, guest data, inventory, pricing, and leverage databases for valuable customer trend insights.

Graduate Research Assistant - Summer 2023

Website design from here