I'm a PhD student in the Computer Science department at the University of Minnesota. My advisor is Professor Arindam Banerjee.


My research interests lie primarily in the intersection of machine learning and natural language processing. In particular, I'm interested in deep generative and probabilistic graphical models, normalizing flows, approximate inference, and information retrieval.

Gradient Boosted Flows
"What I cannot create, I do not understand."
-Richard Feynman

Deep Generative Models

Deep generative models (DGM) can simulate novel and plausible data (e.g. images or text) making them one of the most intriguing directions in machine learning. My research on Gradient Boosted Flows (GBF) increases the flexibility of the latent space learned by normalizing flow-based DGMs. GBF allows practioners to trade additional time training for better performance.

DAP Topic Model
A topic model for temporal, multi-author text data.

Graphical Models & Inference

Probabilistic graphical models (PGM) describe relationships between observed data and hidden variables. I discovered novel PGMs for topic modeling on texts written by multiple authors over time (e.g. blogs), and recommendation systems incorporating text (e.g. movie reviews).
Scaling PGMs to large datasets is challenging. I developed inference algorithms for scaling complex PGMs, with difficult non-conjugate terms, to massive billion-word corpora.

Latent Personas in Journals
Topics extracted from complex multi-author health journals.

Information Retrieval

My research has lead to new methods that can discover common "health journeys" experienced by patients and caregivers during serious health crises. I also have research experience in information retrieval in the health and legal domains, and with a variety of data types.



  • R. Giaquinto and Arindam Banerjee. Gradient Boosted Flows.
    [PDF] [Code]
  • R. Giaquinto and TC. Lu. Structuring Discussions for Collaborative Forecasting.
  • C. E. Smith, Z. Levonian, R. Giaquinto, H. Ma, G. Lein-Mcdonough, Z. Li, and S. Yarosh. "I Cannot Do All of this Alone": Pinpointing Socio-Technical Opportunities for Instrumental Support on Cancer Journeys


  • R. Giaquinto and A. Banerjee. DAPPER: Scaling Dynamic Author Persona Topic Model to Billion Word Corpora. In ICDM 2018.
    [PDF] [Code]
  • R. Giaquinto and A. Banerjee. Topic modeling on health journals with regularized variational inference. In AAAI, 2018.
    [PDF] [Code] [Lightning Talk] [Poster]
  • R. Bjarnadottir, S. Maganti, M. J. Kreitzer, M. Mathiason, R. Giaquinto, and K. Monsen. Discovering the value of the omaha system for knowledge representation and data extraction in health intelligence. In AAAI Joint Workshop on Health Intelligence (W3PHIAI), 2018.


  • H. Ma, C. E. Smith, L. He, S. Narayanan, R. Giaquinto, R. Evans, L. Hanson, and S. Yarosh. Write for life: Persisting in online health communities through expressive writing and social support. Proceedings of the ACM on Human-Computer Interaction (CSCW), 1:73:1–73:24, 2017.


References available on request.

About Me

Me and the mountainsBefore pursuing my PhD, I attended St. Olaf College where I studied mathematics and statistics, and met my wife Lindsey. We have a big, lovable golden retriever named Charlie.

Family and Travel

My wife and I enjoy traveling. Even during graduate school, we make plans every year to see the world. Below are a few of my favorite snapshots of us.


loppetIn my free time I love being active and outdoors, even in winter! Every year I race in the American Birkebeiner and other cross country skiing marathons. During the warmer months you'll find me hiking with my family and eating ice cream.


Please feel free to contact me over email if you want to discuss research, need contact details for my references, or have any questions!
Email: my last name dot ra at gmail dot com