Research

My research is in data science and machine learning, mostly for sequential and time-series data: finding patterns in sequences, building models that forecast what comes next, and turning raw measurements into reliable information. The data I investigate have changed over the years, from protein and DNA sequences to eye-tracking scanpaths, weather, climate, financial markets, and air quality sensors. The underlying questions have stayed the same. Before academia I spent over a decade as a senior software engineer at Rupprecht & Patashnick (R&P), a company that made real-time aerosol measurement instruments. My interests in data science and signal processing, and in filtering, cleaning, modeling, and visualizing data, all come from my early days at R&P.

Nearly all of my work has been (and continues to be) done with undergraduates. See Research Students for the students I’ve worked with, and Publications for papers and posters.

Sequential and time-series modeling

This is where my research began, and decades later, it remains a focal point.

  • Biological sequences. My PhD work with Chittibabu Guda at the University at Albany and my early research at Bucknell developed n-gram and probabilistic methods for protein and DNA sequences: predicting where proteins act in a cell (ngLOC), classifying proteins, de novo genome assembly, phylogenetics, and protein contact maps. See Bioinformatics & Sequence Learning.
  • Forecasting. With Taehwan Kim ’20, deep echo state networks for chaotic time-series prediction (honors thesis, published in Neural Computing & Applications), and with Jason Hammett ’17, data mining for weather forecasting in the Susquehanna Valley.
  • Sequences of events and behavior. With A. Murph and Abby Flynt, comparing event sequences with non-uniform time intervals (Sequential Analysis); scanpath pattern mining in eye-tracking data with Vanessa Troiani at Geisinger (Eye-Tracking Analytics); and statistical word prediction for children with autism who use AAC devices (Word Prediction for AAC).
  • Handwriting. With Bhagawat Acharya ’20, Katie Faull and Carrie Pirmann, handwriting recognition for transcribing Moravian memoirs.

Sensor data and environmental measurement

  • Air quality in the Coal Region. With Doug Collins, Dabrina Dutcher and Shaunna Barnhart, a low-cost sensor network across Mount Carmel, Kulpmont and Shamokin, funded by the EPA Region 3 Thriving Communities Grantmaking program. I’m the project’s data scientist and quality assurance manager. Samuel Tuffour ’28 leads the server-side data pipeline, and he and I are building a research dashboard and a public air-quality website. See this post and the Bucknell News story.
  • Aerosol software. With Dabrina Dutcher and Tim Raymond, LILAC, open software for modeling water uptake by aerosol particles, from an NSF RUI collaboration on which I was a co-PI.
  • Sensors and chemistry. With Ryan Koes ’26, K. Wakabayashi and Alan Marchiori, deep learning on voltammetry and sensor data to characterize coffee (Deep Learning for Coffee Analysis).

Efficient, local AI

Recent student projects share one goal: reducing the computing power, and so the energy, that AI models need, without giving up capability.

  • TTT-Code: test-time training, which lets small, locally run language models keep learning as they read long contexts.
  • Energy per Token: measuring how much energy an AI response actually uses on accessible hardware, and which choices save the most.

Extended reality

  • Gaze Lab (in development): a hands-on Unity lab for eye-tracked headsets, covering gaze interaction, foveated rendering, fixations, heatmaps and scanpaths.
  • Earlier student work used VR to study attention in children with autism (with Vanessa Troiani, Geisinger) and to prototype cardiac surgical training (VR for Healthcare Training, with Dr. Tyler Wallen, Geisinger).
  • My course Design & Development in XR is where many of these projects begin.
Back to top