AI/ML Engineer

AI/ML Engineer

**Job Title: AI/ML Engineer** **Domain:** Healthcare / Medicaid Provider Management (MES / MMIS) **Cloud:** Microsoft Azure **Duration:** 6 Months **About the Role:** We are hiring an AI/ML Engineer to design, build, and operate assistive AI/ML capabilities for a regulated provider enrollment and management platform. The product supports provider enrollment, screening assist, document processing, guided intake, conversational support (chat/FAQ), and triage signals for Medicaid-style programs. AI/ML on this program is assistive Authoritative enrollment and screening decisions remain with deterministic business rules and human review. You will focus on production-grade AI features with strong governance, explainability, auditability, and HIPAA-aligned controls. **What You Will Do (Core Ownership):** - Design and implement conversational AI / RAG (provider FAQ, guided enrollment chat, policy-grounded answers) using Azure AI Foundry / Azure OpenAI, Copilot Studio, and vector search (Azure AI Search / embeddings). - Build document intelligence pipelines (OCR, classification, form pre-fill / Smart-Start patterns) with Azure AI Document Intelligence, integrated into portal and backend services. - Implement AI governance: prompt/version control, grounding and citations, hallucination controls, PII/PHI handling, human-in-the-loop checkpoints, and decision provenance suitable for audits and appeals. - Establish MLOps: model/prompt registry, evaluation harnesses, CI/CD for AI assets, monitoring (quality, latency, cost), and environment promotion (Dev → Test → UAT → Prod). - Integrate AI services with the application stack (e.g., Power Platform, APIs / APIM, containerized services on AKS) using secure, least-privilege patterns. - Define measurable acceptance criteria for AI features (accuracy, grounding rate, latency, cost, exception-queue rates) and iterate with product/BA partners. - Optionally contribute assistive triage / scoring signals on Azure ML where models feed staff review—not as the authority of record. ** Qualifications:** - 4+ years in AI/ML engineering or applied ML in production systems. - Hands-on with Azure AI: Azure OpenAI / AI Foundry, Azure ML, Azure AI Search, and/or Azure AI Document Intelligence. - Strong experience building RAG systems (chunking, embeddings, retrieval evaluation, grounding, citation, safe refusal patterns). - Proficiency in Python for AI/ML services; comfort consuming/producing REST APIs. - Practical MLOps experience: versioning, automated evaluation, monitoring, and secure cloud deployment. - Clear understanding of assistive vs authoritative AI in regulated workflows; ability to design human-in-the-loop systems. - Working knowledge of HIPAA (or equivalent regulated-data) practices: least privilege, secrets management, PHI/PII handling in AI pipelines. - Ability to turn product into testable AI acceptance criteria and ship iteratively. **Nice to Have:** - Experience with Microsoft Power Platform AI patterns (Copilot Studio, adapters/connectors, Dataverse integration). - Exposure to rules engines / DMN (e.g., Drools or similar) and how ML scores feed decision tables without becoming the decision authority. - Entity resolution, fuzzy/phonetic matching, or deduplication assist patterns. - Prior work in Medicaid / MMIS / MES, provider enrollment, credentialing, or other CMS-regulated healthcare systems. - Experience supporting responsible AI / GenAI disclosure documentation for public-sector or regulated programs (supporting Architecture/Proposal—not owning RFP authorship). - Familiarity with Kubernetes/AKS, API gateways, and Azure network isolation for AI workloads. - Domain exposure (as examples only—not ownership) such as: - Screening vendor strategy — evaluating aggregator / CVO / sanctions feeds and integration patterns via an enterprise service bus. - DMN / business rules — ACA categorical risk tiers, appeal-defensible decision tables, rule versioning. - Network adequacy — spatial coverage analytics, geo tooling, or adequacy reporting feeds. - Multi-state productization — configuring state-specific adapters while keeping a shared AI platform core. **Success in the First 6–12 Months:** - Production-ready assistive chat/RAG + FAQ with grounding, audit logging, and safe fallbacks. - Reliable document OCR / extraction path with measurable accuracy and clear exception handling. - Documented AI governance model (authority boundary with rules + humans, provenance, evaluation gates) accepted by Architecture and Security. - Operable MLOps baseline on Azure (registry, promotion path, monitoring, cost controls). **Tech Environment (Illustrative):** - **Cloud:** Azure Commercial; HIPAA BAA; FedRAMP-authorized services where applicable - **AI / ML:** Azure AI Foundry / OpenAI, Copilot Studio, Azure ML, Document Intelligence, AI Search - **App / Integration:** Power Platform, APIM, AKS, secure adapters - **Authoritative Decisioning:** Rules engine (DMN) + human review (owned by rules/platform teams) - **Collaboration:** GitHub / Azure DevOps **Soft Skills:** - Communicates clearly with architects, BAs, security, and engineering partners. - Pragmatic: ships governed, measurable AI—not science projects. - Comfortable collaborating across integration and rules teams without needing to own their domains. - Documents decisions so they are audit-ready. **Education:** Bachelor’s or Master’s in Computer Science, Data Science, Machine Learning, or equivalent practical experience.

Originally posted on Himalayas