CREPES.AI PI - Pilot partnerships now open for Medicare & Medicaid PI teams
Request a Pilot →Stop Medicare & Medicaid fraud faster - with AI that explains every flag, not just raises one.
AI-Powered Healthcare Program Integrity · by CORMAC Health Data Analytics
Investigators are overwhelmed. Existing tools flag thousands of claims with no explanation. CREPES.AI PI changes that - combining ensemble machine learning, clinical AI, and plain-English rationales into audit-ready fraud intelligence that works inside your existing CMS workflows.
CREPES.AI PI runs a multi-layer analysis pipeline - so your investigators spend time acting on insights, not hunting for them.
Part B, DME & Hospice LDS files normalized across provider, beneficiary & procedure.
Multiple ML models run in parallel - catching statistical outliers and billing pattern abuse.
Reasoning AI verifies that diagnoses, procedures & patient demographics actually align.
Ensemble neural network assigns a 0–100 fraud risk score to every provider - ranked for action.
Plain-English rationale generated per claim. Audit-ready export packet in one click.
Every feature translates directly into investigator outcomes - faster decisions, stronger cases, fewer false positives.
Multiple anomaly detection models run in parallel and are synthesized into a single ranked fraud risk score across all providers - so your team knows exactly where to look first.
Every fraud score comes with a feature-level breakdown showing which claim attributes drove the flag - making results transparent, defensible, and ready for audit or legal referral.
Detects clinically impossible or implausible combinations - a prostatectomy billed for a female patient, procedures appearing before the relevant diagnosis, duplicate invasive procedures within days.
CREPES.AI PI is designed around the workflows of real federal health program integrity teams - not a generic AI platform retrofitted for healthcare.
Flag high-risk claims before payment is issued - reducing improper payments at the source rather than pursuing costly post-payment recovery.
Rank providers by fraud risk score to focus limited audit resources on the highest-risk billers - not random sampling. Every referral comes with a documented, defensible rationale.
Apply the same ensemble ML and clinical AI framework to Medicaid claims - detecting billing anomalies, upcoding patterns, and impossible clinical scenarios across state data.
Generate complete investigator export packets - fraud risk scores, claim-level SHAP breakdowns, clinical inconsistency flags - ready for referral to OIG or DOJ without additional manual work.
Plain-English AI rationales and SHAP feature breakdowns translate technical fraud signals into clear, explainable evidence - supporting criminal and civil prosecution of healthcare fraud.
Extend anomaly detection to Part D prescribing patterns and Medicare Advantage encounter data - identifying drug fraud, kickback patterns, and plan-level billing manipulation.
We're running pilot partnerships with PI teams across CMS, MACs, and state Medicaid programs. Let's talk about your specific use case.
Request a Pilot PartnershipCREPES.AI PI stands on years of real CMS work - tested against real Medicare claims data, backed by federal research.
The core ML framework was developed and recognized in the CMS Crushing Fraud Chili Cook-Off - tested against real Medicare Part B, DME, and Hospice claims data.
Federally supported NSF-DOJ-CMS Fraud in Public Healthcare ML research directly informs the anomaly detection architecture.
CORMAC has managed Medicare and Medicaid enterprise data since 2002 - IDR, NCH, MSIS, CWF, Part A/B/C/D, and Medicaid claims.
Anomaly detection built by data scientists who understand CMS clinical and billing patterns - not generic ML applied to healthcare data.
CREPES.AI PI is architected for FISMA Moderate ATO and FedRAMP compliance - not retrofitted. Your data stays where it belongs.
AWS GovCloud-aligned architecture. FedRAMP Moderate ATO roadmap built into the MVP delivery plan from Increment 2 onward. No third-party data sharing - all processing within authorized environments.
Medicare and Medicaid LDS claims data processed under existing CMS data use agreements. No PII exposure in model outputs - fraud scoring operates on de-identified claim attributes.
System security plan (SSP) developed in parallel with MVP. Controls mapped to NIST SP 800-53. Continuous monitoring and incident response procedures included from launch.
CORMAC is ISO 9001:2015 and ISO 20000-1:2018 certified, CMMI Level 3 appraised, SBA WOSB and HUBZone certified - meeting federal contractor compliance requirements.
API-first architecture compatible with CMS adjudication and program integrity workflows - pre-payment, post-payment, and investigative referral pipelines.
We're actively seeking pilot partners in CMS Program Integrity, MAC operations, UPIC, State Medicaid, and DOJ Health Care Fraud. Let's talk about your specific use case.
Contact Us to Start a Pilot