Solutions Architect - AI

Solutions Architect - AI

Location: Indianapolis, IN

Position Summary

The Solutions Architect - AI owns the technical design of how enterprise business knowledge is captured, structured, and served to AI systems, so they answer business questions accurately. The role designs the domain context model, the context repository, and the evaluation harness that measures AI accuracy before and after context is added. It also designs the agentic automation and playbook that letst RADcube onboard new data domains faster than the first. The architect sets technical direction for a team that includes a techno-functional architect, a data and AI engineer, a validation engineer, and a business analyst, and presents progress to client sponsors every week.

Key Responsibilities

Context Architecture and Domain Modeling

• Define the Domain Context Model: the context elements and minimum standards that make a data product "context complete."
• Design the domain ontology (business concepts, relationships, hierarchies, synonyms) and map it to data products and the client's knowledge graph and semantic layer.
• Run a rapid baseline of existing metadata, definitions, and data classification (PII, GxP-regulated, internal, public), and close gaps that block business context.
• Set up ontology ownership, review, and change processes so context stays current as the business changes.

Business Context Repository and Integration

• Architect the Business Context Repository in a client-approved environment, starting simple and hardening it to client security and data standards.
• Integrate the repository with the client's context injection layer, metadata registry, catalogs, and graph, without rebuilding components the client already owns.
• Align access controls to dataset classification.
• Deliver source code, configuration, and runbooks to the client repository.

AI Evaluation and Accuracy

• Design the evaluation harness: ground-truth question and answer sets, an evaluation runner that tests AI answers with and without domain context, and grading by numeric match, query-result comparison, and rubric.
• Build the failure-mode taxonomy (hallucination, stale data, misread schema, wrong join or aggregation, missing business context, persona mismatch) and the scoring rubric.
• Use Claude or another approved LLM to generate and score answers, then iterate on context and re-test to measure improvement.
• Report before and after accuracy by dataset, question type, and failure mode, and rate each data product as AI-ready or not, with remediation.

Agentic Automation and Scale

• Design agentic workflows, orchestration patterns, and human-in-the-loop controls that draft, review, and refresh context at scale.
• Automate intake from existing sources (metadata registry, catalogs, workshop outputs, BI reports, usage logs) and feedback loops from the questions users ask.
• Build scheduled refresh and staleness flags so context does not drift.
• Author the domain onboarding playbook, effort model, and prioritized roadmap for onboarding additional domains across a large data estate.

Client Leadership and Knowledge Transfer

• Serve as the technical lead with client program leads, data owners, and sponsors; confirm scope, success criteria, and priorities.
• Lead weekly demonstrations and incremental releases aligned to the client's release calendar.
• Guide business analysts in turning ambiguous business asks into structured, testable questions.
• Run knowledge transfer sessions and produce handover documentation.

Requirements

Qualifications

• Bachelor's degree in computer science, Data Science, Information Systems, or a related field, or equivalent experience.
• 8+ years in data architecture or solutions architecture, including 3+ years designing AI or LLM solutions for production use.
• Hands-on experience with retrieval-augmented generation, context or prompt engineering, and LLM evaluation (ground-truth sets, automated grading, rubric scoring).
• Experience designing ontologies, semantic layers, or knowledge graphs, and working with business glossaries and metadata catalogs.
• Strong SQL and data modeling skills; able to verify join paths, grain, and aggregation logic.
• Proficiency in Python for evaluation of runners, pipelines, and automation.
• Experience with agentic frameworks and orchestration, with human-in-the-loop review design.
• Working knowledge of data classification and governance for PII and regulated data.
• Clear communicator who can run weekly demos for senior sponsors and translate business questions into testable requirements.
• Preferred: Life sciences or pharmaceutical data experience, including GxP-regulated data. Experience with Claude or other enterprise LLM platforms. Graph database experience (for example Neo4j or RDF/OWL). Cloud certification (AWS, Azure, or GCP). Active client system credentials for immediate mobilization.