AI/ML Engineer

Role: AI/ML Engineer (2 openings) 

Duration: CTH 

Location: Remote 

Job Description — AI/ML Engineer

Domain: Healthcare / Medicaid Provider Management (MES / MMIS)
Cloud: Microsoft Azure

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 only. Authoritative enrollment and screening decisions remain with deterministic business rules + 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.

Required 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 requirements 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 required 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)

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.