Graph ML & Relationship Discovery · SVB

01 / OVERVIEW

Context & problem

Worked on machine-learning solutions for relationship discovery and text categorization during a Data Science & Machine Learning internship at Silicon Valley Bank.

Sparse relationship graphs can hide useful investor-company connections, while unstructured investment information needs categorization before it can support downstream analysis.

02 / SYSTEM ARCHITECTURE

How the pieces connect.

03 / INDIVIDUAL CONTRIBUTION

What I worked on.

• Researched graph neural networks for investor-company relationship graphs.
• Applied vector-search-based link discovery.
• Built text-analysis workflows to classify unstructured investment information.

04 / IMPLEMENTATION

Engineering details.

Sparse relationship graphs can hide useful investor-company connections, while unstructured investment information needs categorization before it can support downstream analysis.

• Combined graph-based relationship modeling with vector-search discovery.
• Classified investment information into 25 categories for downstream analysis.

05 / RESULTS & VALIDATION

What the work supports.

• The relationship-discovery workflow identified approximately 30% more missing investor connections.
• The text workflow organized data into 25 categories.

Validation details are limited to evidence stated in the resume; no unreported datasets, accuracy claims, or benchmark comparisons are added.