Invention of the Year Awards

This approach leverages oscillating link strengths in artificial neural networks, mimicking the rhythmic biomechanical activity observed in neural synapses. By coordinating these oscillations, the network adapts to different contexts and rapidly identifies changes in data patterns. This method allows for unsupervised learning, enabling the network to predict complex system dynamics - even in unseen or non-stationary environments - without requiring predefined contextual tokens. The
algorithm's generalist nature could revolutionize cognitive AI models.

Inventors:

• Wolfgang Losert, College of Computer, Mathematical, and Natural Sciences - Physics and The Institute for Physical Science & Technology
• Hoony Kang, College of Computer, Mathematical, and Natural Sciences - Physics

Cellular Cloaking is a method that allows cellular anonymity by randomly changing the hardware and software IDs in real time at randomized intervals. This function allows a single mobile device to behave like multiple burner phones without the need to purchase ones from a store. The functions performed by the cellular and ID portions of the device are replaced and replicated using a dedicated chip. When these IDs are replaced randomly in firmware, they allow a mobile device to behave as if it was being updated with a new cellular number and device ID at random times. This method also allows the device to use multiple networks and cellular numbers simultaneously without dropping calls or data connections.

Inventors:

• Ray Boatwright, Applied Research Laboratory for Intelligence and Security

The method allows users to objectively measure possible bias in annotations. It uses a generative AI model to create deepfakes that adjust features such as skin color and hairstyle of people in images preserving original expressions. This changes the annotator's perception of sensitive traits such as race and gender of the person. If an annotation for the same person is different before and after the adjustment, it means that there may be bias. Experiments found that the method can
effectively measure bias in annotations, and improved individual fairness by 33%. In addition, using this method to correct the annotations and then retraining the model can also help reduce the bias of the model when making predictions. This methodology offers a framework for addressing biases in pain assessment, criminal justice, education, and employment.

Inventors:

• Siva Viswanathan, Robert H. Smith School of Business - Decision, Operations & Information Technologies
• Balaji Padmanabhan, Robert H. Smith School of Business - Decision, Operations & Information Technologies
• Yizhi Liu,  Robert H. Smith School of Business - Decision, Operations & Information Technologies

RNAnneal introduces a groundbreaking two-stage computational pipeline to predict optimized RNA ensembles. The first stage generates thousands of hypotheses about potential RNA tertiary structures using existing methods. The second stage employs a molecular dynamics-integrated generative artificial intelligence model that scores these hypotheses based on thermodynamics. This method is unique as it does not require pretraining on existing structures, ensuring higher accuracy and efficiency. RNAnneal addresses this challenge by introducing a novel two-stage computational pipeline, improving the precision of RNA structure prediction and enhancing the drug discovery process. RNAnneal was tested on three RNAs from the Critical Assessment of Structure Prediction (CASP15) competition. For each of the RNAs tested, the scoring stage of RNAnneal selected the optimal hypothesis showing 100% prediction accuracy.

Inventors:

• Pratyush Tiwary, College of Computer, Mathematical, and Natural Sciences - Institute for Health Computing
• Lukas Herrong, College of Computer, Mathematical, and Natural Sciences - Institute for Health Computing

It is known that focal weakening of the cornea is the main driver of keratoconus progression and occurs before any shape changes occur in the cornea (i.e. before any clinical symptoms of degraded vision). This invention features motion tracking Brillouin microscopy (MTB), a Brillouin
microscope that measures mechanical properties guided by rapid 3D structural imaging OCT to avoid motion artifacts. It can take biomechanical measurements of the focal area and can diagnose
subclinical keratoconus unlike any previous technology. MTB enabled the capture of high-quality mechanical mapping in vivo even while the subject was breathing normally.

Inventors:

• Giuliano Scarcelli, A. James Clark School of Engineering - Fischell Department of Bioengineering
• James Randleman, The Cleveland Clinic Foundation

The invention is an anti-fungal hydrogel composition comprising a polymer, a carrier, and an effective amount of an anti-fungal agent that can be delivered intranasally It provides a treatment that is both effective and easy to manage. The hydrogel can be applied to the nasal cavity using a syringe at room temperature. It works at just below body temperature (28°C) and can be removed by the body through coughing or nasal discharge. The hydrogel is made using Pluronic F127 mixed in a salt-based solution like saline or water and combined with Amphotericin B – an antifungal drug – in a range of concentrations. Amphotericin B can also be combined with the hydrogel in a different form (e.g. encased in liposomes) to improve its delivery. The hydrogel can also be used with hyaluronic acid, which that helps things stick to surfaces, without interfering with the hydrogel's properties.

Inventors:

• Katharina Maisel, A. James Clark School of Engineering - Fischell Department of Bioengineering
• Kaitlyn Sadtler, National Institutes of Health
• Devon Hartigan, National Institutes of Health

NILE is an innovative technology that integrates near-critical CO2 for the liquefaction and extraction of biomass and municipal organic waste. The process dewaters the biomass using supercritical CO2 and then liquefies and extracts biocrude with significantly reduced oxygen and metal
content. This method produces biocrude that is highly compatible with existing petroleum refining infrastructure, specifically for hydrotreating, while maintaining low acidity, stable viscosity, and improved fuel quality.

Inventors:

• Ashwani Gupta, A. James Clark School of Engineering - Mechanical Engineering
• Kiran Raj Gouad Burra, A. James Clark School of Engineering - Mechanical Engineering

The invention combines micro-encapsulated phase change material (MEPCM) slurries with specialized microchannel geometries, such as helical or herringbone structures, to enhance cooling performance. MEPCM slurries exhibit over three times the thermal capacity within their melting range, enabling effective heat flux management. Using liquid metals like Galinstan or dielectric coolants as base fluids, the structured channels promote 3D transverse mixing, optimizing the cooling potential of MEPCM particles. This results in a compact, stable, and high-capacity thermal management system. Additionally, by implementing dielectric-based cooling, this technology has tremendous potential with heterogeneously integrated 3D chips.

Inventors:

• Damena Agonafer, A. James Clark School of Engineering - Mechanical Engineering
• Vivek Manepalli, A. James Clark School of Engineering - Mechanical Engineering

The invention combines an electron beam source with a remote plasma source. By using reactive oxygen and chlorine-based neutrals, low-damage and selective Ruthenium (Ru) etching can be enabled and controlled by an electron beam. The etching behavior and selectivity to other materials can be precisely tuned using the electron beam and remote plasma source parameters (e.g. high selectivity over tantalum). This method establishes line-of-sight ion-bombardment free processing
for Ru etching, which outperforms conventional plasma etching in both efficiency, spatial control and substrate preservation.

Inventors:

• Gottlieb Oehrlein, A. James Clark School of Engineering - Materials Science & Engineering, Institute for Research in Electronics and Applied Physics
• Yudong Li, A. James Clark School of Engineering - Materials Science & Engineering

DIY is an 8-week classroom program for 3rd, 4th, and 5th grade students that has two components. The first is an individualized guided reflection via an online tool that depicts characters making difficult decisions about inclusion and exclusion in everyday settings like the school, park, or
playground. The decisions are related to things such as gender, race, ethnicity, immigrant status, and wealth status. The second part of the program is facilitated group discussions with all students sitting in a circle, led by a trained teacher. The group discusses the online peer scenarios in terms of students' evaluations, reflections, and interpretations, in addition to discussing personal experiences of social exclusion and solutions. DIY stands out from its social emotional learning (SEL)
competitors by focusing on inclusivity and bias reduction, and addressing factors like exclusion, victimization, and harassment that contribute to the academic achievement gap and stress and anxiety for many students.

Inventors:

• Melanie Killen, College of Education - Human Development and Quantitative Methodology
• Lauren Elenbaas, College of Education - Human Development and Quantitative Methodology
• Michael T. Rizzo, College of Education - Human Development and Quantitative Methodology

Through machine learning and generative AI, users are transported into
virtual galleries and art spaces. Users can walk through a virtual painting,
watch a sculpture take shape in real-time, or even interact with digital art
in ways that were previously unimaginable. By focusing on the digitization
of works from underrepresented artists and artist communities, and
targeting university-based museum collections to digitize, Myseum
presents a solution to distinct yet related challenges. Finally, Mysuem is
an opportunity to resolve the issues related to secondary and tertiary
markets for art, creating new revenue streams from artists from all walks
of life.

Inventors:

• Jordana Moore Saggese, College of Arts and Humanities - Art History & Archaeology; The Driskell Center

An online educational module designed to teach students in any discipline about AI and information literacy. Integrated directly into existing course spaces, it offers an engaging and structured learning experience by including interactive content such as videos, quizzes, and practical exercises to give learners the chance to practice these skills for themselves. The module explains the mechanics and world impacts of generative AI and provides a cognitive framework for evaluating and fact-checking AI outputs using lateral reading skills. The module also provides guidance on citing AI-generated work and using AI tools effectively and creatively. With a focus on academic integrity and critical thinking, it equips students with the knowledge to responsibly navigate AI information ecosystems.

Inventors:

• Mona Thompson, Teaching & Learning Transformation Center; College of Education - Teaching and Learning, Policy and Leadership
• Benjamin Shaw,  University Libraries
• Daria Yocco, University Libraries
• Katie Shilton, College of Information
• Hal Daumé, College of Computer, Mathematical, and Natural Sciences - Computer Science, Institute for Advanced Computer Studies; College of Arts and Humanities - Language Science Center