Skip to content
View astoreyai's full-sized avatar

Block or report astoreyai

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
astoreyai/README.md

Aaron W. Storey

AI Research Engineer | Trustworthy AI Evaluation | Model Selection for Language-Based Decisions | Responsible AI

My in-progress doctoral research at Clarkson University, advised by Dr. Jeanna Matthews, asks when a text classifier is adequate for a specific decision rather than simply best on a benchmark, and whether additional inference computation improves its decisions enough to justify the cost. Requirements come from published deployment practice, and candidate models are compared against a fixed reference with explicit uncertainty bounds and measured cost. No evaluation results are published yet.

I also build research software for multi-agent systems, AI evaluation, and computer vision. I am a PhD candidate in Computer Science at Clarkson University and an independent AI research engineer at Kymera Systems LLC, based in San Antonio, Texas.

Portfolio · Resume · Academic CV · LinkedIn · ORCID · Google Scholar

Current research

  • Question: deciding when a language classifier is adequate for a specific decision and whether extra inference computation earns its cost, testing candidate models against requirements from published deployment practice with explicit uncertainty.
  • Planned methods: non-inferiority testing against a fixed reference, requirement verdicts from one-sided lower confidence bounds, measured inference cost, probability calibration, and selective prediction and deferral.
  • Worked case: detecting subjectively biased wording in Wikipedia sentences to support human editors, using the WIKIBIAS benchmark.
  • Status: a proposed doctoral research direction. No results or publications from it have been published yet.

Selected software

Project Focus
circle Go research-collaboration framework with a knowledge graph, shared memory, artifact tracking, review gates, and 22 MCP tools over git.
ofiqpy Python implementation of the BSI OFIQ v1.1.0 profile: 27 face-image-quality components and a unified score, model integrity checks, and C++ comparison workflows. PyPI.
okf-tools Validation, linting, migration, and indexing for Open Knowledge Format bundles, including relationship and semantic checks.
lunar-terrain-sidecar Lunar terrain authoring for Godot robotics simulation using LOLA/PGDA terrain, ephemeris-derived illumination, and provenance. Software archive.
swinvit-face-verification Common evaluation interface for SwinFace and ArcFace encoders over public benchmark data, with integrity checks and documented limitations.
medicaid-kg Knowledge-graph and geospatial exploration of the 227-million-row HHS Medicaid provider-spending dataset using DuckDB, NetworkX, and agent-assisted querying.
agent-keyring Local credential broker for AI agents on Linux: a loopback proxy injects real credentials so keys stay out of environment files, logs, and model context. Public snapshot with a threat model and security audit.
hostdesk Go Linux service desk that discovers systemd services, containers, sockets, and Tailscale routes, with guarded start, stop, and restart. Release v0.1.0.

OpenFIQA Studio is a design-stage scientific environment for biometric quality and verification research. The repository contains its architecture, scientific schemas, and design records.

Selected publications

Publication record updated November 1, 2026.

Published

  • Ali Kia, Aaron W. Storey, and Masudul H. Imtiaz. Advanced Hardware Security on Embedded Processors: A 2026 Systematic Review. Electronics, 15(5), 1135 (2026). Published March 9, 2026. DOI.

Preprint

  • Yash Sukhdeve, Aaron W. Storey, and Masudul H. Imtiaz. Which Acoustic Features Survive Compression? Tracker-Estimated First-Formant Statistics for Codec-Robust Audio Deepfake Detection. arXiv:2611.13258 [cs.SD], November 9, 2026. Not peer reviewed.

Accepted and presented

  • Aaron W. Storey and Masudul H. Imtiaz. A Three-Channel Audit of the Biometric Evidence Chain. IEEE AIBThings 2026. Presented September 6, 2026; proceedings publication pending.
  • Aaron W. Storey, Ali Kia, and Masudul H. Imtiaz. LLM on the Edge: A 2026 Perspective. IEEE AIBThings 2026. Presented September 6, 2026; proceedings publication pending.
  • Aaron W. Storey, Ajan Ahmed, Yash Sukhdeve, Masudul H. Imtiaz, and Sumona Mondal. Nonparametric Template Persistence Curves for Child Face Recognition via Beta Generalized Additive Models. IEEE UEMCON 2026. Presented October 11, 2026.

Talks and seminars

  • Behavioral Control Surfaces in Large Language Models. Invited seminar, Department of Electrical and Computer Engineering, Clarkson University, January 22, 2026. Slides (PDF).
  • Navigating Worlds Without GPS: Off-World Autonomous Navigation. Three-session public research seminar series, Clarkson University, June 18, June 25, and July 2, 2026. Dead reckoning and SLAM, illumination as a navigation instrument, and autonomous construction on mutable terrain.

Background

  • Clarkson University: PhD Candidate in Computer Science, advised by Jeanna Matthews; member, Explainability & Transparency Group, since September 2026.
  • Kymera Systems LLC: Independent AI Research Engineer, August 2025 to present. Multi-agent prototypes and verification workflows.
  • Clarkson University, AVHBAC Laboratory: Graduate Research Assistant, Autonomous Systems and AI, June 2026 to September 2026. Simulation and evaluation for GNSS-denied navigation, perception, and sensor fusion.
  • Clarkson University, CITeR: Graduate Research Assistant, Trustworthy AI and Biometrics, June 2025 to May 2026. Biometric measurement, explainable AI, and face-image quality.
  • Education: PhD study at Clarkson University from January 2025; M.S. in Artificial Intelligence, Maryville University (2024); B.S. in Information Technology, American Military University (2017).
  • Teaching: Nine guest lectures for EE622 Advanced Biometrics; co-creator of the free 100 Days of Machine Learning curriculum.

Before graduate research, I served in the United States Army, August 2002 to January 2018, across medical operations, technical training, information management, and deployed communications, including Defense Health Agency assignments. I completed my undergraduate IT degree during service.

Tools: Python, Go, R, SQL, PyTorch, Hugging Face Transformers, scikit-learn, FastAPI, DuckDB, Docker, Linux, Git, ROS2, OpenCV, and Godot.

For research collaboration or AI research engineering opportunities: storeyaw@clarkson.edu.

Pinned Loading

  1. claude-skills claude-skills Public

    Personal collection of Claude Code skills for enhanced AI-assisted workflows

    Python 5 1

  2. money-talks money-talks Public

    Five-class trading and investing curriculum in 100 Jupyter notebooks for Google Colab.

    Jupyter Notebook

  3. 100daysofml/100daysofml.github.io 100daysofml/100daysofml.github.io Public

    100 Days of Machine Learning Challenge

    Jupyter Notebook 86 21

  4. astoreyai astoreyai Public

    GitHub profile README.

    HTML