Principal AI Security Researcher · NVIDIA

Ferhat Erata

I combine machine learning with formal verification to make AI systems safer and more secure.

Role
Principal AI Security Researcher, NVIDIA
Location
Remote, NJ
Doctorate
Yale University, 2025
Interests
AI Security, Formal Verification
ferhat@ieee.org
Google Scholar
GitHub
/in/ferhaterata
0000-0001-6305-4266
Biography

Principal AI Security Researcher at NVIDIA AI Security Research. Previously, I was an Applied Scientist at AWS Agentic AI, in the Automated Reasoning Group and Kiro Science. I built neurosymbolic systems combining LLMs with formal verification, with contributions to Bedrock Guardrails, Mainframe Modernization, and Kiro.

PhD, Computer Science, Yale (2025). Advised by Ruzica Piskac, with Jakub Szefer as co-advisor.

Before Yale: co-founder of an R&D startup in Turkey; led EUREKA/ITEA consortia with Airbus, Daimler, and Ford (ModelWriter, ASSUME, XIVT), building open-source formal-methods tools (AlloyInEcore, Tarski) deployed in industrial settings.

§ II News
§ III Writing

Occasional essays on neurosymbolic AI, automated reasoning, and the craft of research. All essays →

§ IV Experience
  1. —
    Present

    Principal AI Security Researcher

    NVIDIA — AI Security Research Remote, NJ

    Research on securing AI and agentic systems through automated reasoning, formal verification, and neurosymbolic methods.

  2. Feb 2025 —

    Applied Scientist

    AWS Agentic AI — Automated Reasoning Group & Kiro Science New York, NY

    Built production neurosymbolic safety systems for Bedrock Guardrails, AWS Mainframe Modernization, and Kiro.

  3. Sep 2019 —
    Feb 2025

    Research Assistant & Teaching Fellow

    Yale University New Haven, CT

    Doctoral research on learning-based automated reasoning and quantum computing security. Teaching Fellow for Operating Systems and Database Systems (Avi Silberschatz); recipient of Yale CS Distinguished Teaching Award (2023–2024).

  4. Jun 2022 —
    Dec 2024

    Applied Scientist Intern (three internships)

    AWS — Automated Reasoning Group New York, NY

    Contributed to initial techniques behind Automated Reasoning checks for Bedrock Guardrails. Designed a scheduler framework in Rust for randomized and model-based testing of distributed AWS services, integrated into S3's model-based testing pipeline.

  5. Dec 2008 —
    Aug 2019

    Research Engineer & Co-founder

    UNIT Research & Development Ltd. Ege University, Turkey

    National consortium leader for EUREKA/ITEA projects (ModelWriter, ASSUME, XIVT) with Airbus, Daimler, and Ford. Led the development of formal-methods tools deployed in industrial settings. Publications at ACM FSE, ASE.

§ Education
  • PhD in Computer Science

    Yale University New Haven, CT

  • MSc & MPhil in Computer Science

    Yale University New Haven, CT

  • MSc in Information Technologies

    Ege University Izmir, Turkey

  • BSc in Computer Engineering & Industrial Engineering (double major)

    Dokuz Eylul University Izmir, Turkey

§ V Publications

Peer-reviewed papers and preprints. Filter by year or topic.

Filter