PhD Candidate | Explainable AI Researcher | Opening the Black Box
Making AI interpretable, trustworthy, and auditable.
| Area | Focus |
|---|---|
| XAI Methods | Attribution, counterfactuals, faithful explanations |
| Transformer Interpretability | Logic-gated modules, attention probes, mechanistic analysis |
| Adversarial ML | Robustness testing, perturbation analysis, model brittleness |
| Applications | Biometrics, LLMs, vision models, safety-critical systems |
PhD Thesis: Developing falsifiable attribution methods for explainable AI - systems that can be empirically validated and audited.
Proposal: May 2026 @ Clarkson University | Walk: May 2027
| Project | Description | Status |
|---|---|---|
| Beta Regression Framework | Statistical framework for bounded biometric performance in child face recognition | IEEE T-BIOM (R&R) |
| Agentic AI Seminar | Personas & affective prompting as behavioral control surfaces | Clarkson Seminar |
| Goblin Forge | Multi-agent CLI orchestrator for Linux with tmux isolation | Active Dev |
| SIFTER | NASA Space Apps 2024: ML seismic detection for moon/marsquakes | NASA Hackathon |
| 100 Days of ML | Complete 35-lesson curriculum: Python basics to XGBoost | |
| AIRA | AI Research Assistant: PRISMA 2020 + NIH-compliant research automation | 11 agents, 5 domains |
| Money Talks | Trading & investing education: 100 notebooks, 5 classes | Complete |
| EE622: Biometrics Transformers | 10-week graduate course: ViT for fingerprint, face, gait, speaker, ECG | Clarkson University |
- Dissertation: Falsifiable attribution methods for explainable AI in face verification
- IEEE T-IFS (Submitted): Falsifiable attribution via counterfactual score prediction
- IEEE T-BIOM (R&R): Beta regression for bounded biometric performance
- IEEE T-BIOM (Submitted): Persistence curves for child face recognition
- IEEE T-BIOM (Submitted): Individual heterogeneity in template degradation
- IEEE T-BIOM (Submitted): XAI policy/standards alignment (AI Act, GDPR, Daubert)
- IEEE T-BIOM (Under Review): Dual-backbone ViTs for pediatric age estimation
- IEEE TDSC (Submitted): Prompt injection analysis via Gandalf challenge
- ACM CSUR (Submitted): FIQA systematic review 2018-2026
- NeSy 2026 (Submitted): NSPL neuro-symbolic prompt logic
- MDPI Electronics (Published): Hardware security systematic review
- Founder & CEO @ Kymera Systems: AI orchestration platform
- Portfolio: astoreyai.github.io
- LinkedIn: linkedin.com/in/astoreyai
- ORCID: 0009-0009-5560-0015
- Email: [email protected]
"Make explainability an engineering discipline: measurable faithfulness, human factors by design, and governance that withstands audits."