Graphletter
Open-source GRC product that compares organizational evidence with control objectives across 81 supported frameworks.
LLM extractionSecure Controls FrameworkEvidence pipelineSource-cited reasoning
Peter Boaz — Personal site
Product Security / GRC
Peter Boaz
I work in product security for medical devices. Most of the job is helping R&D teams make good security calls before release, especially when the policy and the product do not fit together neatly. I also build tools for work that should not depend on somebody repeating the same manual steps. Graphletter is one of them.
Now
- Guide US R&D teams through product-security milestones at ZEISS; own cybersecurity premarket readiness for the US portfolio's FDA 510(k) submissions.
- Expanded Graphletter to 81 supported frameworks and added exact-header validation after a source upgrade exposed a systemic mapping error.
- Shipped a rule-correctness grading gate in BarPlaybook after validating it against the human-graded golden set.
- Earned CRISC (Certified in Risk and Information Systems Control), ISACA.
Featured
Tools for making expert judgment measurable, reviewable, and easier to use.
BarPlaybook
AI-graded bar-exam essay platform. Every scoring change is checked against a 32-essay human-graded golden set (MAE 2.58; 84.4% within 5 points).
Evaluation harness (MAE / within-5)Golden set + calibration loopLLM scoring with deterministic post-processingHuman-grounded evaluation