Aman Sharma Principal Enterprise Architect AI/ML | Independent Researcher | AI Governance & Safety in Healthcare I design AI/ML platforms for healthcare and research what breaks when LLM systems operate in clinical settings. 18 years in IT, with the last several years focused on enterprise architecture for AI/ML platforms and multi-agent AI governance. Currently building AI/ML reference architecture across Databricks, Azure AI Foundry, and Snowflake for a health plan covering 6M+ members. Independently researching multi-agent LLM safety, emergent misinformation, and governance frameworks for regulated industries.
Open Source AI Architecture Enterprise Patterns 18 enterprise AI architecture patterns with 228 interactive data flow visualizations. Covers unified AI gateway, RAG for regulated data, multi-agent safety gates, compliance-aware routing, governance-as-architecture, contamination-resistant pipelines, FinOps, security, observability, and more. Zero install required. Explore live demos GAIF Governance Observatory Open-source governance toolkit for the Governed AI Architecture Framework (GAIF). Six interactive CLI tools covering governance maturity assessment, risk scoring, compliance gap analysis, and architecture fitness evaluation. 25 passing unit tests. Contamination Percolation Research codebase for measuring how errors propagate between AI models in multi-agent pipelines. Introduces the T1PR (Tier-1 Percolation Rate) diagnostic and demonstrates the gap inversion effect across DBRX, Claude, Llama, and Gemini model families.
Research My independent research focuses on what breaks when LLM systems operate in healthcare. All work is solo-authored. PaperStatusVenueEMG (Emergent Misinformation Genesis)PreprintedTargeting NeurIPS 2026 D&BPHI-GUARD (Compliance-Aware LLM Routing)Under reviewIEEE JBHIContamPerc (Contamination Percolation)Under reviewIEEE AccessMedMI-Bench (Clinical MCQ Benchmark)Under reviewJMIR AIGEG (Governance Effectiveness Gap)Targetingnpj Digital MedicineTEMPORAL-MED (Temporal Consistency)TargetingJMIR AIGNC (Governance Non-Compositionality)TargetingNeurIPS 2026 GAIF Framework: GAIF-4 v1.5 defines four quantitative metrics for AI governance health: EMR (Emergent Misinformation Rate), T1PR (Tier-1 Percolation Rate), CFR (Compliance Failure Rate), GDR (Governance Decay Rate).
Preprints & Publications
EMG — Zenodo PHI-GUARD — TechRxiv GAIF — Zenodo | SSRN GAIF-4 v1.5 — Zenodo
Professional
IEEE Big Data 2026 Program Committee (Industry & Government Track) IEEE Access Reviewer NIST Public Comments: CAISI Framework, NCCoE AI Guidelines MCP Enterprise Interest Group — Healthcare & Compliance Use Cases Champion MS in progress, Colorado Technical University
Connect
LinkedIn — Newsletter: AI Architecture & Governance ORCID Email: Aman_sharma007@yahoo.com