Machine learning inference models and NLP algorithms
For harvesting data from external sources, enhancing entities with additional attributes, inferring missing attributes, standardizing unstructured text fields, and classifying documents
System of neural networks and genetic algorithms for breaking homomorphic ciphers
System and method for delivery and integration of multi-sourced targeted information on remote display devices
Software/Hardware framework for delivering visual images and interactive applications to one or more remote displays located in residential and business settings
Allows individual user customization of content and preferences
Patent application acquired by a private firm
Arguably reads on several high-profile products that have since hit the market
System and method for intelligent automated remote management of electromechanical devices
Patented system consisting of an on-premise wireless network of controllers and sensors with two-way control and data transmission to a data center, and with SaaS management console for controlling, analyzing, and reporting
System for intelligent automated response to line frequency and voltage disturbances by a central control point communication with a plurality of load control devices
Patented system for automatically detecting and responding to power disturbances on the electric power grid that result in under/over frequency and under/over voltage events
Plant pest infection prediction model
Framework and algorithm for encoding empirical expert knowledge of plant pest infection propensities into an automated alerting system
Computational algorithm for inline calculation of crossing-rate functions
Algorithm for efficiently calculating extreme value crossing-rate functions from digital signals inline with the data acquisition process
Selected for, and presented in, the plenary session of the Eighth International Conference on Wind Engineering
Novel approach for developing site-specific wind climate models
Probabilistic wind climate models are required for the design of large structures, such as skyscrapers, long-span bridges, and sports stadia
R&D for this approach was proposed to, and funded by, the National Science Foundation, and published in this final report
Provided distinct advantages over the previous industry standard SOTA approach, which was to select wind speed records from the single nearest airport available and extrapolate to the subject site
These sources are notorious for malfunctions of (mechanical) anemometers, missing data, poor maintenance, and poor data due to bad physical placement of anemometers
Developed and demonstrated a new approach that avoided all these deficiencies, by using more robust data from (non-mechanical) barometers at multiple locations across a region to reconstruct the macro-scale pressure field, and then using first principles of physics to determine gradient level and site-specific wind speed time series
Results were remarkably successful and demonstrably superior to the standard approach
System for rapid custom analysis of a residential structure for wind damage risk
Commercial studies of individual residential structures for extreme wind damage risk involved the sequential application of several procedures involving manual CAD preparation of structural details, then fed into analytical models and Monte-Carlo simulations
This required 1-2 weeks to complete any single study and report, which made it infeasible to provide the service as a commercial product to clients that included State Farm Insurance, and State of Florida
Developed process and software to reduce the entire process (going from input request to final auto-generated custom home-owner report) down to 2 hours with minimal manual steps
Resonant frequency damping for pneumatic pressure sensors
Long pneumatic tubes are used in wind-tunnel studies to transmit pressures at locations on a wind-tunnel model to sensors located outside the wind-tunnel
These suffer from resonant frequency effects
Developed an approach for optimized placement of damper devices by using direct mathematical analogies to a similar problem being dealt with in the utility industry for electrical resonances in transmission lines
Algorithm for inferring user preferences for residential thermostat settings
Developed algorithm for deriving optimal configuration of thermostat control scheduling by analysis of user behavior in adjusting thermostat over time
For implementation in PowerMand's DreamWatts(r) Energy Management System
Product and algorithm pre-dated products such as the Nest Learning Thermostat (now owned by Google)
Algorithm for expanding magic squares in size
Developed a mathematical algorithm for expanding a given magic square in size
Used this to create a very large magic square that was recognized by Guinness World Records for breaking the previous record by over 1000X
Guinness then, in direct response to breaking the record, removed the category from their book because it had now been "reduced to an algorithmic process" that, in their view, "removed the competitive incentive for others to break the record"
Algorithm for extending magic squares to higher dimensions
Developed a mathematical algorithm to take a magic square in 2 dimensions and extend it into higher dimensions while maintaining its mathematical 'magic' properties
Framework for interactive educational physics simulations
Developed a software framework for rapidly developing interactive simulations to demonstrate various physics scenarios, which was then used to create several such simulations for IBM's World Book Encyclopedia
Framework for rapid development of interactive marketing-support web applications
Developed a software system for the rapid implementation and hosting of web applications in support of short-term promotional campaigns being provided as a marketing service to client companies
This system provided a high-level declarative language so that these interactive websites could be fully implemented by non-programmers
