(Submitted) Real-time Attack Verification using Embedded Network-geometry (RAVEN)

Collaborative proposal with Brown University and Rice University focused on real time detection and mitigation of adversarial manipulation of AI models during training and at inference.

(Awaiting Funding) Learning Unknown Mechanisms in Electro-optics for Native Spectra (LUMENS)

Project focused on identification of unknown phenomena that limits the utility and performance of existing tradecraft for spectral image processing.

(Funded) Agile Scheduling and Tasking with Retrospective-data Analysis (ASTRA)

Transforming the scheduling and allocation of resources through the development of an end-to-end ecosystem and evaluation harness. Metric definition and synthetic data generation from historical data to inform and stress-test system robustness and efficiency for mission use.

(Funded) Validation & Integrity of Spectral Techniques and Algorithms (VISTA)

Development of mission-relevant T&E for AI for hyperspectral data analysis to accelerate the integration and operationalization of fundamental research efforts. Supporting overall AI model accreditation activities for stakeholders.

(Funded) Uncertainty Estimation and Robustness to Out-of-Distribution Data in AI Models

Development of techniques for performing model-agnostic uncertainty estimation including capturing data and performance drift without access to model internals.

(Funded) Foundations for the Future: Scientific Progress in AI Research and Knowledge (SPARK)

Strategic internal investment focused on foundations and understanding of existing AI technologies, novel architectures, and alternative learning paradigms. Efforts within the investment include Multi-objective Optimization and Learning Techniques, Diffusion & Distributions, Mathematically-Inspired Neural Design, Static Capability Analysis and Learning Evaluations, and Scalable Compute Approaches for Inference-time Reasoning.

(Funded) Unsupervised Deep Clustering of Seismic Time Series Data at Volcanoes

NSF funded research through my UTEP joint appointment focused on the adaptation of existing AI technologies and development of new architectures for clustering and event discovery in seismic time series data.

(Funded) Physics Informed Compression and Denoising for Hyperspectral Signal Processing

Internal seeding effort focused revisiting enduring challenges in hyperspectral signal processing including atmospheric correction that preserves underlying physics knowledge.

(Funded) Leveraging a Universal Topological Feature Space for Multi-Modality Data Fusion

Exploration of the effectiveness and robustness of multi-modal fusion of topological features through “multi-spectral” persistence images. Emphasis placed on operational settings where post-deployment data collects may be incomplete due to sensor failures.

(Funded) Artificial Intelligence for Advanced Manufacturing

Development of AI techniques to accelerate the R&D cycle in advanced manufacturing using the Shear-Assisted Processing and Extrusion (ShAPE) solid-phase processing technology. Building on the process, structure, and property paradigm we built AI models to understand the bidirectional relationships between processing parameters and resulting material properties through composition and invertible neural networks.

(Funded) Topological Data Analysis for Naval Optics Challenges

Demonstrated utility of topological data analysis techniques for questions of interest in support of the Optical Sciences Division of the US Naval Research Lab. Success was shown for performing motion-track disambiguation taking a dynamical systems+TDA approach.

(Funded) Data-driven Approaches for Predictive Maintenance of Oil-wetted Machinery

Feature extraction from distributions of detected particles of different types and sizes from existing tradecraft to identify indicators of failures in oil-wetted machinery for the Navy.

(Funded) Detection, Characterization, and Attribution of Deception in Documents using Topology

Assessment of learned representations of text-based data, pre-LLMs, to reveal patterns related to deception, misinformation, and fraudulent research papers. Seedling internal investment was focused on answering “Can you find a circle in circular logic?”