Experience

University of Southern Mississippi

• Research direction: trust-calibrated reinforcement learning for adversarially robust multimodal agents. Advised by Dr. Rabab Abdelfattah.
• Current manuscript: Learning When To Trust — training-time calibration for reinforcement learning under adversarial observation corruption (manuscript in preparation).
• Developing a learned reliability estimator for sequential decision-making under corrupted, shifted, and adversarial observations; experiments published in the open through TraceVox Research.

Research overview →

Deep Anchor Solutions

• Led development of reproducible 2D axisymmetric finite-element workflows for Deeply Embedded Ring Anchor performance in sand, studying soil plugging, load capacity, displacement, and soil–structure interaction.
• Automated parametric studies linking wave-energy-converter size, mooring loads, and anchor sizing, generating 5,000+ simulation runs for surrogate-modeling research.
• Produced simulation-to-ML datasets, connecting scientific computing outputs to machine-learning research, in collaboration with engineering teams.

TraceVox (TraceVox Research & TraceVox AI)

• Built TraceVox Research, an open environment for reproducible RL and trustworthy-AI experiments with recorded per-timestep traces, exact replays, and provenance-carrying research bundles; distributed publicly via PyPI.
• Engineered TraceVox AI, a production AI observability and evaluation platform: tracing, scored evaluations, safety monitoring, and agent/LLM infrastructure.

Software →

Applied machine learning experience

• Several years of applied ML and engineering experience: ML pipelines, LLM applications and guardrails, geospatial ML, and cloud-native systems (AWS, GCP, Azure).
• Led development of end-to-end 3D mineral-exploration ML pipelines, benchmarking CNNs and GNNs against Random Forest and XGBoost with spatial cross-validation, calibration analysis, and feature diagnostics.
• Sole-authored a systematic LLM safety evaluation (625-prompt adversarial prompt-injection benchmark, SSRN preprint, 2026).
• Peer-reviewed geospatial ML research: sinkhole susceptibility via machine learning (Applied Computing and Geosciences, 2025).

Publications →

Prior career

• 3D geological modeling and data-intensive workflows for resource projects — reasoning under uncertainty, rigorous documentation, and cross-disciplinary delivery that now inform my approach to ML research.