Kora Dey

(2 Year UC Leadership Excellence through Advanced Degrees Program)

• Designed and deployed a domain-specific ChatGPT-based research assistant using OpenAI, LangChain, and FAISS, trained on academic publications, statistical texts, and university datasets to support undergraduate research.
• Developed an intuitive natural language interface that allows users to generate code, interpret statistical models, and receive real-time guidance on reproducibility and research methodology.
• Fine-tuned model performance using Python, Jupyter, and Hugging Face Transformers, with extensive prompt engineering to align responses with disciplinary standards.
• Currently leading efforts to institutionalize the tool across multiple UCR departments, aiming to democratize access to advanced AI research support and enhance student-faculty collaboration.

• Designed end-to-end data pipelines in Python and R to automate preprocessing, normalization, and visualization of high-throughput experimental datasets.
• Led advanced analysis of high-dimensional reaction data using unsupervised learning techniques including PCA, hierarchical clustering, and k-means, enabling identification of hidden chemical behavior across experimental conditions.
• Developed modular, reproducible frameworks using pandas, ggplot2, and dplyr, supporting batch experimentation and accelerating time-to-insight for new chemical workflows.
• Took initiative in modernizing the lab’s data ecosystem by enforcing FAIR principles, coordinating scalable data practices across projects, and improving readiness for publication and cross-collaboration.

• Assisted in a public health research initiative focused on understanding disparities in food marketing and access across California.
• Collected, reviewed, and organized large datasets from retail food environments, including store audits and community-level health indicators.
• Ensured data consistency and accuracy across multiple regions, supporting the lab’s broader statistical analysis and reporting.
• Contributed to a research brief used by health agencies to address nutrition equity in underserved areas.

• Developed high-fidelity epidemic simulations using Poisson multigraph SEIR models in R and Python to examine transmission across structured population networks.
• Conducted network-level statistical inference to identify dominant group interactions and evaluated the impact of dynamic intervention strategies.
• Created custom visualizations using ggplot2, igraph, and matplotlib to summarize complex transmission pathways.
• Delivered an internal report that modeled targeted outbreak containment strategies in heterogeneous populations.

• Trained a convolutional neural network using TensorFlow and Keras to classify malignant and benign mammographic findings, with the goal of improving early breast cancer detection
• Preprocessed high-dimensional imaging datasets by normalizing, augmenting, and resizing inputs to enhance model accuracy and reduce diagnostic bias
• Tuned sensitivity and specificity to benchmark model performance against clinical standards, exploring its use as a decision support tool for radiologists
• Presented results at the Synopsys Science and Technology Fair, where the project was recognized for innovation in medical AI, with mentorship from faculty across multiple institutions

• Designed a behavior classification model using scikit-learn and trained it on criminological datasets to identify psychological risk profiles, learning how AI can uncover hidden patterns in human decision-making.
• Used SHAP to visualize feature importance and interpret predictions, gaining hands-on experience with cutting edge explainable AI tools and the ethical challenges they address.
• Explored algorithmic fairness by benchmarking model outputs across sensitive demographic groups, and proposed mitigation strategies for bias and representation gaps.
• developed understanding of applied AI in high-impact social systems, culminating in a national symposium presentation on the role of machine learning in criminal justice and ethical AI deployment.

• Built full-stack data pipelines using Python, SQL, and Flask to process and model structured datasets, applying supervised learning techniques for real-world predictions
• Automated data cleaning, normalization, and visualization using tools like Plotly and Seaborn to streamline analysis workflows
• Completed accelerated training in Python, Java, JavaScript, and C++, developing cross-functional software applications in collaborative team environments
• Led beginner coding workshops, supported peers in debugging and logic fundamentals, and cultivated a lasting commitment to accessible computing and education equity

• Developed and tuned classification models using scikit-learn, pandas, and NumPy to address real-world business challenges including wine quality prediction and customer lifetime value ranking
• Built end-to-end machine learning pipelines with modular preprocessing, hyperparameter tuning, cross-validation, and automated performance tracking
• Deployed models into production by integrating outputs with internal dashboards and collaborating with engineering teams to ensure seamless, real-time insights for stakeholders
• Strengthened practical knowledge of applied machine learning, focusing on model interpretability, business alignment, and scalable solution design