Category Boundary
Machine-verifiable does not mean machine-authorized. Sovrient produces evidence states that can be independently inspected, reconstructed, challenged, and verified; auditors, IV&V teams, contracting officers, certifying authorities, review boards, and counsel retain authority.
- Not a compliance determination, legal clearance, certification, or authorization layer.
- Not courtroom admissibility doctrine.
- Not a general compliance attestation service.
- Not a generic audit-trail product.
- Not a replacement for third-party assurance or certification.
Layer Model
Activity Layer
Git, Jira, LMS, telemetry, AIS, documents, AI agents, and operational workflows generate events and outputs.
Evidence Layer
Sovrient converts events and outputs into bounded evidence: source custody, manifests, receipts, supportability states, blocked claims, and replay conditions.
Authority Layer
Auditors, IV&V teams, contracting officers, certification bodies, counsel, and review boards evaluate evidence and make acceptance decisions.
Core Competencies
Replayable Decision Verification
SSEJ-complete records for machine-mediated systems: what the system saw, selected, excluded, and justified under a bounded decision scope.
Deterministic Replay
Replayable evidence artifacts with fixed inputs, declared methods, and fail-closed verification gates for independent recomputation.
Maritime AIS Evidence
Bounded AIS review packages with observation-gap review queues, manual resolutions, proof handles, and a governed replay dashboard over public NOAA AIS data.
Trigger-Safe Payload Design
Policy-bound trigger payload schemas that separate evidence generation from downstream routing, hold, or settlement action.
Prime / Government Fit
Sovrient does not compete with sensor providers, collection networks, autonomy stacks, or generic dashboards. The company fits as machine-verifiable evidence infrastructure and a replayable decision verification layer that sits above collection and beside mission or review systems.
- Supports maritime domain awareness review, anomaly triage, and after-action reconstruction.
- Supports unmanned route accountability, operator override replay, and incident-support workflows.
- Supports future advisory and hold payloads where machine action must be traceable to bounded evidence.
GovCon Evidence
GovCon Evidence is the government-contracting application of the evidence layer: the same source custody, supportability states, and replay conditions applied to the capture, pursuit, and proposal lifecycle. It screens which pursuit and staffing assumptions are evidence-supported before bid & proposal dollars are committed. It is a record discipline, not a market-intelligence feed: it does not predict winnability, assert eligibility, give pricing advice, or replace GovWin-style discovery.
Pursuit Evidence Packets
Deterministic SAM.gov pursuit-vector packets over a declared NAICS/notice-type window, with an offset-walk coverage contract (saw-versus-total asserted, fail-closed), per-record decomposed scoring, retrieval sidecars, and OpenTimestamps anchoring. Bounds what was seen and excluded; it is a metadata snapshot, not a fit, incumbent, or award-likelihood claim.
Pre-B&P Evidence Gate
A sealed gate memo that resolves declared pursuit assumptions against the captured public record as byte-bound quotes, returning a four-state support surface (supported, contradicted, silent, out-of-record). Demonstrated against a live federal Sources Sought, surfacing a record-contradicted assumption before proposal spend. The artifact computes states; the human owns the bid / no-bid / partner-path decision.
Labor Evidence Screen
Scaffold that screens staffing and rate assumptions against a firm's own vehicle / LCAT evidence, joined by verbatim quote to the captured solicitation. Runs on the evidence owner's side; rate values stay sealed locally and never cross the boundary export, enforced by a rate-leak gate and the externally anchored hash tree. Demonstrated on synthetic vehicle data; not pricing advice or a staffing commitment.
Counterparty Evidence Atlas
Public-safe map from public demand signals to PEO/program context, counterparty archetypes, evidence burden, Sovrient lane fit, blocked claims, and operator next steps. It tracks archetype alignment and evidence posture; it does not infer private pursuit intent, name a company as interested, or convert public signals into bid / no-bid authority.
Each motion chains to the next as one custody line — pursuit packet to gate memo to labor screen to counterparty atlas — so a reviewer can replay discovery, selection, exclusion, and archetype-aligned attention from committed evidence states. GovCon Evidence is a vertical wedge of the same governance reconstruction category; it does not assert a separate methodology.
JADC2-Compatible Boundary
DoD describes JADC2 in terms of helping forces sense, make sense, and act across domains. Sovrient is architecturally aligned with the make sense portion of that model: we govern how observed inputs become trusted, bounded state and how resulting decisions become replayable, reviewable evidence artifacts. We do not claim to be a full JADC2 command-and-control fabric, sensor network, or action layer. We claim a narrower role: replayable decision verification and accountable evidence surfaces that can fit inside broader JADC2-style architectures.
Selected Relevant Work
Maritime AIS Governance Lane
Published a full Hampton Roads AIS slice with canonical bundle, observation-gap review, manual queue, resolution ledger, and replay dashboard. Current public state: 25 queue entries, 25 recorded resolutions, 0 unresolved.
Local-Enclave Governance Benchmark
Demonstrated deterministic governance over a 69,006-node / 69,720-relation memory graph on commodity laptop hardware with replay-exact hash equality, fail-closed behavior, and hostile-reviewed bundle support. This is a bounded proof of local-enclave governance performance and replay integrity, not a claim of universal workload coverage or autonomous decision authority.
Two-Prover Formal Layer
Public theorem-reference and machine-state surfaces now declare two prover families in bounded terms: a certified ACL2 family with live raw frame and ZF gate twin routes, plus a build-verified repo-internal Isabelle/HOL family for session scope, participant authority, and opportunity-brief safety. This is broader independent formal coverage, not a claim that Isabelle twins are public or that every lane is fully theorem-bound.
Operating Specification Twin
Published a document twin with manifest, companion markdown, verification handles, and state receipt under a deterministic normalization path.
Seismic Verification Surface
Maintains a live multi-source verification surface with machine-state discovery, sealed catalog outputs, and bounded admissibility posture.
Website-Safe Proof Block
Sovrient has hostile-reviewed benchmark evidence showing deterministic governance over a 69,006-node memory graph on commodity local hardware. The supported benchmark bundle preserves replay-exact hash equality, fail-closed behavior, contradiction handling, and declared exclusion accounting, and it carries a hybrid trace-signature surface verified offline during review. This is a bounded proof of local-enclave governance performance and replay integrity, not a claim that every operational lane or workload is already covered at the same scale.
Public Proof Points
- Maritime AIS dashboard for replay, queue state, and resolution detail.
- Maritime AIS lane for public scope boundaries and source posture.
- Governance reconstruction as a supporting methodology page for the older cross-lane framing.
- Theorem reference and theorem-reference.json for the bounded two-prover formal story.
- Controlled-review benchmark anchor available on request for the hostile-reviewed local-scale MAS bundle.
- machine-state.json and agent-manifest.json for machine-readable discovery.
GovCon Profile
Primary Classification
NAICS 518210, Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services. Default secondary alignment is 541512 and 541715, with 541360 used only when geophysical analytical work is sold as a discrete line. NAICS 541690 remains an advisory-only optional addendum, not the default public posture.
Business Status
Small Disadvantaged Business. Current public surface intentionally keeps federal identifiers out of the website machine state; SAM, UEI, and CAGE are inserted into controlled leave-behind materials before external submission.
Point of Contact
Email: info@sovrient.com. Website: sovrient.com.
Compliance & Security Posture
CMMC 2.0 Scope
Under current public-data flows, the public evidence lanes are best read as outside the present CMMC Level 2 assessment scope. If a future customer-specific lane processes, stores, transmits, or protects federal CUI, Sovrient would reissue the scope determination and apply the corresponding control and assessment requirements for that lane.
NIST SP 800-171 / FedRAMP
Current public surfaces are compliance-supporting and security-relevant, but they are not presented as a completed NIST SP 800-171 assessment for a CUI-bearing enclave. No FedRAMP authorization or in-process claim is made for this public surface.
Security Reporting
Public vulnerability reporting is published at /.well-known/security.txt. Initial public intake uses info@sovrient.com; engagement-specific secure channels are established separately when required.
Agent Access & Training Use
Verification access is open to independent readers, browsers, and interactive agent fetch tools. Bulk AI training-corpus extraction is a separate permission boundary and may be restricted at the edge unless separately licensed. Current exceptions are declared machine-readably in capabilities.json.
SBOM & Supply Chain
SBOM and supply-chain attestation posture are furnished in controlled review or engagement-specific packets. They are not implied by the public website alone.
AI RMF-Aligned Evidence Crosswalk
Sovrient should be read as AI RMF-aligned evidence infrastructure and AI RMF-aligned evidence packaging. This page does not claim AI RMF compliance, NIST certification, FedRAMP authorization, or generic GRC replacement. The narrow claim is evidence packaging: a qualified reviewer can inspect the bundle, verify the evidence chain, apply the same policy pack, and reach the same release verdict without trusting the original vendor narrative.
AI RMF is voluntary in form; federal AI acquisition evidence is becoming contractually load-bearing in practice. Sovrient packages that evidence so it can be independently reviewed and re-judged under the buyer's policy boundary.
Govern
Policy packs, authority boundaries, SSEJ records, scope declarations, and residual-risk decisions make governance reviewable. Sovrient supports the evidence substrate underneath CAIO, legal, acquisition, and risk-acceptance authority; it does not replace those authorities.
Map
Evidence graphs, component lineage, AIBOM-style inventory, source declarations, model or tool manifests, and excluded-input records map what the release saw, selected, excluded, and justified.
Measure
TEVV outputs, deterministic replay, tamper tests, monitoring records, incident records, and fail-closed verification gates let reviewers measure whether release evidence satisfies the declared policy pack.
Manage
Release verdicts, hold states, incident escalation, reviewer notes, anchor receipts, and successor manifests preserve risk treatment decisions and make later re-judgment possible.
Federal AI Acquisition Fit
M-25-21 Boundary
OMB M-25-21 frames federal AI around innovation, governance, and public trust. Sovrient's public fit is the evidence layer that helps reviewers inspect how a release was tested, monitored, documented, and accepted under a declared policy boundary.
M-25-22 Boundary
OMB M-25-22 makes acquisition evidence concrete: agencies need fit-for-purpose AI, performance validation, risk tracking, and cross-functional review. Sovrient supplies independently re-judgeable evidence packages for those conversations, not procurement authority or legal approval.
Critical Infrastructure Profile Wedge
NIST's AI RMF Trustworthy AI in Critical Infrastructure Profile effort is the standards lane closest to Sovrient's architecture: AI bills of materials, traceable rationales, deterministic behavior, fail-safe operation, graceful degradation, and rigorous TEVV. Sovrient should treat that profile as a standards-entry path, not as a current certification claim.
Review Artifact Boundary
Desktop or browser LLM reviews are useful adversarial review evidence, but they are not deterministic execution traces unless prompts, outputs, model/version claims, timestamps, source hashes, and transcript hashes are captured into the evidence chain.
Near-Term NERV Artifacts
- NERV-001: buyer one-pager translating Sovrient into AI RMF / OMB evidence language.
- NERV-003: reproducibility demo pack with publication manifest, policy pack, evidence graph, timestamp sidecar, known-good bundle, tampered bundle, verifier, and expected outputs.
- NERV-002: NIST Critical Infrastructure Profile comment or Community of Interest submission once the one-pager and demo pack are coherent.
A review-ready ANATOP standards twin for the NIST AI RMF Playbook is published at /standards/nist/ai-rmf/playbook/1.0/, with source hashes, JSON-LD outcomes, an evidence crosswalk, and a validation report. It is a derived navigation and evidence-packaging artifact, not an official NIST artifact or an AI RMF compliance claim.
Procurement Posture
Current Sales Motion
Current public positioning fits best as machine-verifiable evidence infrastructure and replayable decision verification for primes, government technical evaluators, and product-adjacent teams. This page does not make a prime-ready claim.
Vehicles & Identifiers
No GSA schedule, SEWP, OASIS+, CIO-SP3, or other contract vehicle claim is made on this public surface. UEI, CAGE, SAM status, and any vehicle posture are handled through controlled procurement materials.
Export-Control Posture
Open public-data lanes are published as UNCLASSIFIED // OPEN SOURCE. Export-control analysis is handled per customer-specific data, destination, and integration scope; no blanket ITAR or EAR claim is made here.
Authority Boundary
- No settlement authority.
- No underwriting, reserve, or valuation opinion.
- No legal adjudication.
- No claim that a bounded public-data lane alone proves intent, fault, or full-world state.
Next Procurement Artifact
This page is the public capability surface. Contracting conversations that proceed beyond technical evaluation should use the procurement packet and controlled leave-behind materials for identifiers, vehicle posture, and customer-specific security or compliance questions.
For a shorter prime and government leave-behind, use the teaming brief. For capture-specific framing, use the prime-contractor one-pager. For evaluator-facing reading order, use the procurement evaluator memo. For analyst-facing category framing, use the analyst category note. For the anchor category essay built on the same substrate, use The Negative-Assertion Pattern. For a buyer-run public-substrate integrity check, use Buyer Verification.