AI assurance research and cloud security. I build research software that exposes what an AI-assurance claim establishes, what it assumes, and where it must stop.

Site → cubits11.github.io — every technical claim there links to public evidence or is explicitly marked attested; the evidence ledger carries the bindings.

• CC-Framework — composed guardrail failure as a partial-identification problem: sharp Fréchet–Hoeffding bounds, claim envelopes, receipts, decay semantics. Five-minute version: When Marginals Are Not Enough.
• Ghost-Ark (S2 Lab, Penn State) — a verifier and measurement harness for the provenance limits of AI-governance receipts.
• Assay — early-stage, private: which provenance claims can a verifier actually derive from AWS Nitro Enclave attestation.

Penn State CS ’26, Cybersecurity minor · AWS CCP + AI Practitioner · Philadelphia, PA.

Repositories labeled [HISTORICAL] or [SUPERSEDED] are preserved as records of earlier exploration, not as current claims — some carry pre-discipline language I would not write today. Any repository from 2025 or earlier without a status label should be read the same way. Current work is what's pinned, plus the site above.