Data Science Intern

An AI-enabled product selector that puts product, training and technical knowledge in one place for customer-facing sales representatives and new team members, built and deployed end to end for internal use.

• Users: customer-facing sales representatives who need product details in front of a customer, and new employees who use it for training and technical onboarding.
• Model: a selector that works across multiple data modalities, including PowerPoint decks and video, so an answer can come from wherever the knowledge actually lives.
• Data pipeline: read source data out of SharePoint and built an automated ingestion pipeline, so the data refreshes on its own and the selector stays current without manual updates.
• Deployment: carried the project from raw source data to a deployed application, used internally by sales representatives and new hires.

Data Science Intern

A graph-RAG question-answering system over 20 years of software manuals, built for Applications Engineers.

• Users: Applications Engineers, who need reliable answers from two decades of technical documentation.
• Approach: used LangChain to convert document knowledge into a Neo4j graph database, with open-source and Azure models, so retrieval returns context-rich results.
• Hard cases: supports temporal, vague and reasoning questions over extensive technical documentation.
• Impact: gives engineers scalable search across the full manual archive.

Data Science Intern

Graph neural network for product recommendation, plus research on generative models for new product formulations.

• Recommendation: built a heterogeneous graph and used node classification, with attention to information flow in non-homogeneous graphs, to improve recommendation accuracy.
• Generative research: explored variational autoencoders for novel product formulation on graph structures, and identified how large, dense graphs bias the latent space and hinder stable VAE training.