The Linguistic Security Institute (LSI) is an open-source research initiative dedicated to ensuring AI systems are culturally coherent, institutionally resilient, and ethically verifiable. We audit linguistic bias in AI, develop open-source tools for bias detection, create educational resources for developers and students, and propose governance frameworks for linguistic safety in AI development. Visit LSI →

Directed multilingual digital strategy, expanding organic web traffic by 62% and overseeing cross-functional transcription/translation programs in English, Spanish, and Arabic.

Programming & Tooling

Python (Pandas, NumPy), SQL, Git, GitHub, Bash environment scripting.

AI & NLP Methodologies

LLM Evaluation, Error Analysis, RAG architectures, Prompt Engineering, Zero-Shot Learning, Bias Analysis.

Data & Quality Assurance

Data Annotation, Localization workflows, Gold Label Audits, Rigorous Dataset Validation.

Linguistics & Languages

Dialectology, IPA/SAMPA.
English & Spanish (Native), Arabic (Writing & Linguistic Analysis: Working Proficiency)  Italian (Intermediate).

Arabic Polarization Detection Across 10+ Large Language Models

Second-author and Error Analysis Lead. Investigated dialectal bias, hallucinations, and Clever Hans effects across 10+ models (DeepSeek, Qwen, Gemma, Llama, Fanar). View Paper →

New Announcement Coming Soon

New Announcement Coming Soon

Linguistic Firewall: Polyglot Poisoning

Open-source Python toolkit for detecting multilingual prompt injection and polyglot poisoning vulnerabilities.  An expanded, reproducible version of the demo originally presented in April 2026 will be featured as an official hands-on workshop at DEATHCon BSides on November 14, 2026.

Mitote

RAG text-to-speech prototype supporting Nahuatl-aware Spanish pronunciation for Indigenous language preservation.