Victoria M. Garibay, Ph.D.

The Deep Graph Library Poverty Trap Model (DGL-PTM) is part of an ongoing effort in collaboration with the Netherlands eScience Center to create an Agent Based Model (ABM) framework that supports communication between agents in a network with a heterogeneous population on the scale of millions. Supporting elements from the Deep Graph Library, particularly message-passing, enabled us to leverage matrix operations on GPUs to achieve the necessary capacity.

Role: keeping the project on track towards research goals, balancing the ideas of the PI and the eScience team to keep things progressing, and eventually taking a greater role as a developer when administrative upheaval impeded the team's ability to continue.

Experience gained with: PyTorch, DGL, GitHub collaboration (big one!), software verification (static analysis, model checking, unit testing), HPC job scheduling, API documentation, socioeconomic ABMs

Existing skills exercised: experimental design, technical writing, Python, pandas, NumPy, matplotlib, seaborn

View DGL-PTM on GitHub

View DGL-ABM Generalization Project on GitHub

This project is a comprehensive demonstration of my capabilities in independent project management. It is also a showcase of my skills in the context of developing tools for the formation and visualization of business insights. Inspired by my own difficulty finding accommodation that suited my needs, I set out to create a product that could inform site placement of budget cabinettes near U.S. National Parks and Monuments. I have been consistent about documentation of both the code and the development process. Although it may be very far in the future, one day, I hope to find the investment required to actualize the concept. The first MS PowerBI iteration was completed on schedule in September 2025. A new iteration of the project was envisioned with a fully online dashboard and is currently live (see project log for progress here).

Role: leading the project from inception to completion and ensuring timely delivery of project milestones

Experience gained with: DMAIC principles, PostgreSQL, Power BI, TypeScript, React, Redux, Next.js, Node.js, Reactour

Existing skills exercised: project management, problem framing, product development, database querying, data cleaning, data engineering, dashboard development, UX design, Agile methodology, geospatial analytics, APIs

Cabinette Project Log

View SQL backend & PowerBI iteration on GitHub

View online dashboard APIs & frontend on GitHub

View online dashboard on Vercel!

This project was inspired by a need established when scaling up an agent based model (DGL-PTM, leftmost featured project). The function responsible for modeling agent decisions became a tremendous bottleneck. As a solution, I developed a Multi-Layer Perceptron (MLP) that predicts the outputs of the original Bellman equation approximation with an efficacy that made it a viable replacement.

Role: picking a direction for the replacement function, developing a methodology for selecting a model, conducting the training and testing

Experience gained with: deep learning, algorithmic documentation

Existing skills exercised: PyTorch, experimental design, technical writing, Python, pandas, NumPy, matplotlib, seaborn, HPC job scheduling, LaTeX

View on GitHub

This project from my Ph.D. was one of my first efforts in dashboard development. It was extra gratifying because it was made available to a broader audience. Designed as a simple tool to geographically organize data available to the project which spanned across three countries in East Africa, the dashboard also came to include aggregated summaries of the data,serving as a communication tool for the project team and stakeholders. While the dashboard itself is no longer available and is, as a 2021 project, outdated, it was a great learning experience which allowed me to explore what I now acknowledge as a small subset of the possibilities of interactive visualizations.

Role: picking a platform for the dashboard, determining its scope given the available resources, testing the dashboard with stakeholder feedback, arranging for the app deployment with the help of the Purdue Research Computing team

Experience gained with: dashboard development, flexdashboard, Shiny, dygraphs, leaflet

Existing skills exercised: data structuring, database querying, data cleaning, merging data sources, bias correction, R, Plotly, GIS, dplyr, stakeholder communication

https://app.climate-dashboard.geddes.rcac.purdue.edu/ (no longer active, now links to sad image that remains)

This project features Lenny, a conversational AI assistant designed to help visitors of this website. Lenny served well as a practical way of exploring the possibilities of cloud computing and LLMs (large language models). The assistant is available for use below. Lenny was implemented through the integration of Azure Foundry AI (GPT-4.1-mini) and Azure AI Search services (orchestrated within an Azure Function). The AI Search service indexes documents pertaining to my work and expertise, allowing the assistant to give contextually relevant responses to user queries.

Role: requirements gathering, technology selection, solution architecture, Azure resource management, end-to-end implementation (Python, Azure Functions, AI Search, OpenAI/Foundry), integration of conversational AI and semantic search, deployment, user experience testing, and iterative improvement.

Experience gained with: cloud computing, API development, Azure, conversational AI, LLM integration, semantic search indices, CI/CD automation, managing secrets

Existing skills exercised: python, version control, static analysis, unit testing, project management, UX design, HTML, CSS, JS

Project Management Summary

View on GitHub