In build
Coding copilot
ICD-10-CM / CPT suggestions with NCCI validation. A certified coder stays in the loop. The model does not finalize the claim.
CAC Co-Pilot for certified coders
Clinical notes are rich. Coding workflows are not. Verisource Labs builds CAC Co-Pilot — a Cloud Run assistant that proposes ICD-10-CM / CPT codes for a credentialed coder to accept or override. NotifyLawlor is a DBA only.
Schema-bounded packets from synthetic outpatient charts. A coder reviews what the engine saw.
Triage plus NCCI / Excludes1 checks before anyone commits a code set.
Grok API critic and Vertex agree/disagree/insufficient judge. Stubs until program acceptance.
Human in the loop. Persist the decision. No DFT / FHIR writeback in the MVP.
ICD-10-CM / CPT suggestions with NCCI validation. A certified coder stays in the loop. The model does not finalize the claim.
IAP in front of a CPU service in us-east4. No local GPU on the coder path. Demo charts are synthetic.
AHIMA/ACDIS-style, non-leading documentation queries when the note is ambiguous.
CAC Co-Pilot runs on Cloud Run behind Identity-Aware Proxy. Accept / override rows persist to a store that can move to Cloud SQL or BigQuery. Real charts wait on written BAAs. Hosting an image is not Safe Harbor.
Cloud Run + IAP in us-east4. Vertex is the judge only. Grok API is the clinical critic.
Startup Demo is synthetic. No public page ships real patient data.
Credentialed coder review (CPC / CCS and related). Assistive output, not authoritative claims.
Four decades of large-scale systems integration and distributed infrastructure, including Apex Global Internet Services (AGIS) and national backbone work (ASN 4200). Founder of Verisource Labs, LLC.