Machine-Speed Defense. Sovereign-Grade Evidence.
Chokmah is industrial cybersecurity, functional safety (FuSA), and AI engineering under one sovereign-grade practice. Every claim, cyber or FuSA, ships with a computationally checkable artifact, not an adjective.
Wedge offer · AI custody
We don’t sell a smarter agent. We sell an agent that won’t pretend.
Most AI products sell more capability. Chokmah sells governed capability: AI-assisted workflows with authority edges, evidence boundaries, independent review, and claim gates so consequential work does not launder uncertainty into a deliverable. If it wasn’t verified, we will not dress it up as verified.
Awkward on purpose. Useful after the first confident mistake. This is advisory and workshop practice, not a compliance certificate, and not a claim that models never err.
Draft and prepare inside the request. Sending, buying, deleting, and compliance status stay human-gated.
Observed, inferred, and unverified stay labeled. No costume certainty for the board packet.
Independent critique and executable checks where numbers matter. Green only when the check actually ran.
That block wasn’t a bug. That was the product. Start with a scoping call or a paper-backed study, not a skill catalog tour.
Limited advisory capacity during the defender's window: retained through Feb 2027, ~5–8 hrs/week remaining. Book a call →
Private training on secure SDLC, AI-assisted spec work, and cloud security is available by inquiry. Contact us to discuss scope and fit.
Automated Regulatory Transformation for Critical Infrastructure
The Capability: An automated engine that converts unstructured NERC-CIP cybersecurity requirements into machine-readable OSCAL v1.1.2 component definitions with intelligent NIST SP 800-53 mapping.
Designed to compress unstructured NERC-CIP requirements into machine-readable OSCAL component definitions with an automated test suite, so audit preparation produces a verifiable artifact trail rather than one-off document sprints. Client-specific time savings are measured per engagement (not stated as a universal ratio here).
An agent with standing access and no audit trail is an unmonitored insider. Chokmah extends its audit-first, OSCAL-based traceability practice to agent permissions and actions: a scoped identity, a least-privilege boundary, and a log that survives a federal audit, aligned to IEC 61508 / NIST 800-series requirements. See Governed Capability.
Chokmah architects the co-scaling loop: the objective function, the claim-gate and review checkpoints, and the tooling selection. Your team or operations partner runs continuous AI-assisted threat modeling and remediation without drowning in agent-generated alerts.
Chokmah applies STRIDE, PASTA, and attack-tree methods to agentic AI systems so risk is expressed as measurable exposure, not only red/green checklists. Secure SDLC automation is part of the same engagement pattern.
Anomaly detection, automated triage, context-aware alerting, and controlled LLM integration into defensive workflows. Alert-quality and handoff improvements are measured per engagement, not claimed as a universal percentage.
Context engineering structures retrieval and rules before the prompt layer so models load less noise and reuse cacheable context. Release metrics for Claude Code Playbook v4.4.0 are on the research landing (Headroom 47–92% dynamic context reduction; path-scoped rules 41% overhead reduction; DOI 10.5281/zenodo.20481459), stated as author-reported figures, not client guarantees.
Occasional private workshops on secure SDLC, AI coding assistants, and compliance-oriented automation, scoped per engagement. Contact us to discuss fit.
Computer security, digital forensics, programming/code analysis, risk profiling, and log interpretation, via testimony, independent reports, and court-preparatory consulting under the docket qualification noted above.
Chokmah designs defense-in-depth for SCADA, DCS, and PLC environments using protocol-aware segmentation and IEC 62443 practices so lateral movement inside critical segments is constrained without forcing unnecessary downtime.
Using the January 2026 toolkit, unstructured NERC-CIP requirements become machine-readable OSCAL v1.1.2 component definitions with NIST SP 800-53 mapping, JAMA integration, and 27 automated tests. Regulators and internal audit can inspect definitions plus checks instead of narrative-only write-ups; hour-saved deltas remain engagement-specific when measured.
FuSA (functional safety) oversight covers hazard analysis and logic-level verification for AI-integrated Safety Instrumented Systems (SIS) in high-hazard zones such as battery energy storage (BESS), chemical, and power. Spec-driven checks aim to keep AI-assisted operations from overriding hardware safety constraints.
Roughly 70% of serious vulnerabilities trace to memory-unsafe code, a legacy of C/C++ written when compute was scarce, not a fact of nature. Chokmah applies its spec-driven method here: write the formal behavior spec first, then scope and prioritize which legacy OT/ICS modules justify AI-assisted Rust migration, with an artifact trail proving behavioral equivalence. This is advisory and triage, not a full migration build-out.
We do not sell generic best practices. Chokmah engineers Sovereign Grade systems where every technical artifact is computationally true, legally defensible, and strategically unassailable.
We move security and project management away from opaque "red/green" checklists toward mathematical defensibility. By enforcing Spec-Driven Development, that is, architecting formal specifications (CUE, YAML, Markdown) before code generation; we ensure system logic is defined by rigorous constraints, not chatbot probability. We deliver "Impossibility Proofs" backed by dependency manifests and Monte Carlo simulations, quantifying risk abatement in dollars and weeks, not adjectives.
Competence is only Sovereign Grade if it survives a high-stakes federal audit. We treat configuration management as a safety-critical path, mandating linear history and cryptographic signing to satisfy IEC 61508, UL 1998, and UL 5500. Hardened by NIST SP 800-series alignment, we transform daily logs into a Strategic Audit Trail that pre-answers the scrutiny of federal court experts.
We practice the architectural minimization of information waste. By structuring retrieval paths before the prompt layer, we trade "hallucination" for maximum clarity. We "buff" messy engineering signals into high-resolution executive narratives that leadership can act upon immediately without interpretation.
We act as the "invisible architect" who makes leadership appear high-velocity and proactive. We frame technical implementation as Roadmap Protection and Cross-Functional Enablement, providing the sovereign technical foundation that allows leadership to focus on high-level stakeholder alignment and strategic execution. Cyber-physical convergence, OT/ICS, functional safety, and enterprise cyber under one practice, is not a pitch here. It is already how Chokmah operates.
We maintain a relentless bias toward provable transparency. We identify "binary blobs" or proprietary backdoors as disqualifying risks, pivoting to auditable, open-source solutions to eliminate vendor-managed opacity. Sovereign competence means maintaining full Software Bill of Materials (SBOM) transparency; your security posture should never rely on unverified vendor claims.
Dr. Daniyel Yaacov Bilar brings over 20 years of sovereign-grade operational, research, and academic experience to every engagement:
Dr. Daniyel Yaacov Bilar is Principal of Chokmah LLC. ORCID: https://orcid.org/0000-0002-9040-6914.
An automated engine (delivered January 2026) that converts unstructured NERC-CIP requirements into OSCAL v1.1.2 component definitions with NIST SP 800-53 mapping, JAMA integration, and 27 automated tests.
As stated for Q2 2026–Q1 2027: retained through February 2027 with about 5–8 hours per week remaining.
At https://chokmah.me/research/ with Zenodo DOIs (OSF mirrors).
Governed capability is Chokmah’s name for AI-assisted work under custody: authority edges, evidence boundaries, independent review, and claim gates that refuse false certainty. It is advisory and workshop practice, not a compliance certificate or a claim that models never err. See Governed Capability.
Yes. We treat every agent as an insider requiring identity, least-privilege access, and audit logging, expressed the same way we express NERC-CIP/OSCAL component definitions: machine-readable and independently checkable.
Capacity notice (self-stated): Retained through February 2027. About 5–8 hours/week of remaining capacity.
Initial scoping calls are 30 minutes via Cal.com for projects or cases that fit this profile.
Direct channels: [email protected] | Github: README