I specialize in extracting quantitative detail from large data sets
(click heading to see more)
Taxonomy - Natural Language Processing
Using machine learning, I have developed a language model, based on Google's BERT, to develop custom taxonomies for scholarly publishers and associations.
Hum uses machine Artificial Intelligence (AI) to understand client content, create a unified & weighted taxonomy, and connects the taxonomy to keywords for each individual piece of content. Our process works great with completely untagged data and applies automatically to new content as it is generated.
User Engagement Scoring
At Hum, we've developed a proprietary algorithm to measure and develop user engagement with content over time. How a user interacts with material speaks volumes about their interests and what value an organization can add to their interests. The flow and timing of this process is key to providing value to both the user, content generators, and publishing organizations.
We've partnered with the University of Virginia's school of data science graduate program to build a fractional attribution project at Hum.
It's crucial that any organization or publisher understand a user's journey from first contact to participation in a high-value event if they want to retain users and provide them value. We're using cutting edge machine learning techniques to do all the heavy lifting and give organizations and marketers advice on what motivates their user base.
Submarine Landslides and Slope Stability
Conditions within subsea sediments determine when, where, and how slope failure occurs. Seismic data is an excellent tool for investigating both sediment and pore fluid characteristics. Using data from the ENAM community experiment we are developing a robust picture of the conditions in and around the Cape Fear and Currituck landslides.
By applying modern processing techniques to legacy data, I have been examining the CO2 storage potential of Mesozoic rift basins offshore the US east coast. Detailed prestack waveform inversion sheds insight into basin compositions, and thus the character of the rock infill and volumetric storage potential.
My research goals aim to map diapycnal diffusivity of meso- and sub-mesoscale ocean structures. Using spectral methods applied to carefully processed seismic data I quantify turbulence around internal waves, rough bathymetry, and eddies. Field sites include the Caribbean, South China Sea, and Adriatic.
Using prestack waveform inversion, my research goal is to detail the distribution of methane hydrates and free gas. Then by leveraging detailed velocity information, move toward making quantitative estimates of hydrate concentration, particularly in coarse grained sands. My study areas have been at Blake Ridge and various sites in the Gulf of Mexico.
Working with borehole geophysicists, I am developing and applying methods to estimate rock properties such as pore fluid pressure, porosity, and hydrate concentrations from inverted seismic data. The aim of this research is primarily to address natural hazards along continental margins.