Capability Statement

Core Competencies

• A records engine, protected by ten patent filings. The LEF SI Engine reads the shape of the evidence around a record, not the record alone. Ten U.S. provisional patents pending cover cognitive graph traversal, structural void detection, cross-domain pattern transfer, tesseract composition and cross-engine reasoning over distributed knowledge graphs. The records engine runs part of what they describe; each licensed build carries what its corpus needs.
• Reads what a body of records is missing, and tests each gap. Finding what is missing is common: constraint-discovery tools do it too. What the engine adds is what it does with each gap. It names the pairings of values the rest of the records say should appear, and do not, counts each against every record supplied, and says for each how firmly the records hold it, the evidence on why it holds (never a cause stated as fact), what would change it, and one next step. A person decides; the tool built around the engine does the work that follows.
• Reads a dataset's documentation against its records. Rule by rule: what agrees, what the records contradict, what is documented but never used, and what is written differently, with counts, and whether the documentation, the records, or both explain each gap. A documented missing code changes nothing until the dataset's steward accepts it.
• Reads test records against their test plans. With NHTSA's FMVSS 208 laboratory test procedure written by LEF as requirements, each citing the procedure, the engine checked all 165 of NHTSA's FMVSS 208 barrier tests dated after the procedure for every channel it requires; two lack a required vehicle accelerometer.
• Shows its own limits. Gaps the records refute stay on record as refuted, thin ones are reported as leads, and every run writes down what it checked and what it could not, with the reason.
• Known-answer benchmark. On the FDIC institution directory (27,834 records), against textbook methods LEF built from their papers: all 66 planted gaps found; across eleven reads of 5,000 rows, no absence the full file contradicts (constraint discovery stated 96; an outside open-source method, Metanome's CFDFinder, stated none false, but 19 in all); all 6 kinds of misleading record that can be caught from the file alone. Not an independent evaluation.
• Four working prototypes on public data. Research Articles, Biology, Patents and Public Services (federal records on a public site; state and local records in reports).
• Delivered as a licensed container. Runs in the customer's own environment with the customer's keys, reaches only the hosts it declares, and never connects back to LEF. On 28 October 2026, two LEF builds were sealed and run in Google Cloud Confidential Space (AMD SEV), with the image attested and every engine check passing inside the enclave.

Government Engagement

• Rated Awardable: IARPA Solutions Marketplace, 8-26-0025, Decision Support at the Command Edge (September 2026).
• Favorably evaluated, not funded: AFRL Rome, GeoPEX BAA (white paper, October 2026) and AFAR BAA (white paper).
• Submitted, awaiting decision (August to October 2026): DARPA Strategic Technology Office (abstract, October 2026); DARPA ERIS (resubmitted October 2026); Army Applications Laboratory BAA (white paper, October 2026); Army SBIR (DoW 2026 CSO); Naval Research Laboratory long-range BAA; DOT PHMSA hazardous materials research BAA; AFRL Information Directorate and Human Effectiveness Directorate; DLA research BAA; AFOSR open BAA; IARPA and CDAO Tradewinds marketplace submissions.
• Received: DIA NeedipeDIA (white paper routed to mission owners, September 2026).
• Cornerstone OTA (COTA) consortium member.
• NSF active Principal Investigator registration (NSF ID 0000A8D1S).
• State of Nevada registered Vendor (T29052309), NevadaEPro, GOED-certified Local Emerging Small Business Tier 1.

Differentiators

• Structure, not similarity alone. The engine reads the structure of a body of records as a whole. App. Nos. 63/993,278; 63/993,317; 63/993,979; 63/993,984; 64/002,205; 64/023,988; 64/043,294; 64/045,185; 64/061,710; 64/061,715.
• A gap must survive the records before it is shown. Incomplete collection and a genuine gap can look identical, so every gap is checked against every record supplied before it stands, the run says when the records cannot tell which it is, and it reports what it could not check.

Company Data

• Socio-economic. Small business; nontraditional defense contractor; NV Local Emerging Small Business. No foreign ownership, control or influence; 100% U.S. owned.
• Security posture. NIST SP 800-171 Rev. 2 Basic Assessment posted to SPRS, score 87 of 110, assessment date 6 Nov 2026, enterprise scope, against System Security Plan LEF-SSP-002 with a maintained POA&M. This satisfies DFARS 252.204-7019. CMMC Level 1 self-assessment completed Jun 2026. No facility clearance and no cleared personnel.
• Key staff. Zontonnia Moore, Founder. MS Positive Psychology (ASU 2025); M.Ed., Curriculum and Instruction (Applied Behavior Analysis concentration) (ASU 2019).
• Procurement. Government purchase cards accepted (set up with the buyer at award). Pre-priced engagements under FAR Part 13 micro-purchase (under $15,000) and simplified acquisition ($15,000 to $350,000) thresholds. Licensing priced to scope; rate sheet on request.

Changes to This Statement

This page is always the current version. If you hold an earlier copy, this is what changed and when.

• Version 13, 29 October 2026. The founder’s master’s degree is now stated as the diploma prints it: M.Ed., Curriculum and Instruction (Applied Behavior Analysis concentration). Earlier versions gave it as “MEd Applied Behavior Analysis”.
• Version 12, 28 October 2026. The container sentence now dates the 28 October seal and no longer says “on the current engine”: the sealed builds carry an earlier engine version than the current one.
• Version 11, 28 October 2026. What the engine says about a gap's cause is now stated as the evidence on why the gap holds, never a cause stated as fact. "One engine, different corpora" is removed: the four prototypes run their own engines, and each licensed build carries its own engine, scaled to its corpus. The patent line now says the records engine runs part of what the filings describe. The container line now cites the 26 October 2026 seal on the current engine.
• Version 10, 21 October 2026 (Cornerstone OTA consortium membership added 22 October). Added the read of test records against their test plans (FMVSS 208); dated the container seal to 6 October 2026; added the AFRL Rome evaluations and new submissions; purchase-card wording clarified.
• Version 9, 12 October 2026. Added the read of a dataset's documentation against its records. The gap bullet now says what the engine adds beyond finding what is missing.
• Version 8, 6 October 2026. Rewritten around the records engine as it runs. Removed claims that the engine composes the resolving action, the description of the four prototypes as production deployments, a self-refutation and false-discovery-control claim, and a comparison with named commercial tools. Added the known-answer benchmark, the limits every run reports, a gap's cause "where the records show one" (corrected in version 11), the licensed container sealed in Confidential Space, and the IARPA Awardable rating.
• Version 7, August 2026. The earliest version kept as source. Versions 4 to 6 (May to July 2026) were sent before source copies were kept.