Costa Cassis

Costa builds AI systems that have to survive contact with a regulator, not just a demo. Over more than 20 years he has turned enterprise data into decisions for healthcare payers, providers, the public sector, and global retail, and he has done it as both the executive who sets the strategy and the engineer who ships the code. He builds the teams, platforms, and governance that move AI from a board memo into work people actually use, and when the moment calls for it, he architects and ships the platform himself. He currently leads enterprise AI at Medecision as its first dedicated AI leader, building the function from the ground up across both sides of the business: the population-health and care-management SaaS platform, and the AI advisory practice (formerly Excell Healthcare Advisors, acquired by Medecision in September 2025). He wrote the company's AI governance posture against regulation that already exists rather than theory: human-in-the-loop design, model documentation, and audit-response plans built against Alabama SB 63, California SB 1120, and the UnitedHealthcare nH Predict and Cigna PXDX cases. He defined the AI service offerings the company takes to market (readiness assessment, governance stand-up, strategy and roadmap, and workflow redesign across utilization management, prior auth, care management, and member services), and he drove adoption past engineering: expanding Claude access, turning on Copilot where it was licensed but switched off, and forcing a written AI use policy after the rules were being decided ad hoc in chat threads. He is also the founder of Future Era AI, where he built Lumina, a multi-tenant, AI-native analytics platform that lets a non-technical person ask a question in plain language and get back governed SQL, a chart, and a written narrative, without copying a byte of the customer's data. He architected the federated query engine himself on Trino and SQLAlchemy, answering questions across 17+ source types (Snowflake, BigQuery, Redshift, Databricks, Postgres, and more) by querying each source on demand. The natural-language-to-SQL pipeline runs on a Qdrant semantic layer with Claude and OpenAI behind a provider factory, and a query-time governance engine handles column masking, row-level filters, SQL-parsed lineage, immutable audit logging, and per-tenant credential encryption. Production runs on Azure Container Apps under a single BAA through Microsoft AI Foundry, which makes the platform usable against HIPAA-bound data. He also built Statute Watch, an agent that tracks state AI legislation affecting healthcare payers end to end: pulling the daily legislative digest, updating a curated bill-tracking workbook while preserving human-curated fields, and flagging new bills for review, unattended. Earlier, he was Managing Director of Data Intelligence at NTT DATA, where he directed enterprise data transformation for payer, provider, and public-sector clients, led a 150-plus person global delivery team, and oversaw $15M+ in realized value by integrating AI with revenue-cycle automation, public-sector grants, and care-management innovation. He launched a cloud-first data stack on Snowflake, dbt, and Airflow, decommissioned 25+ legacy systems and mainframes, and led predictive modeling that cut claims-review cycles by 65%. Before that, as Program Director at Anthem, he led a $5.5M enterprise API program to deliver ONC/CMS FHIR-compliant endpoints, and as Director of Solution Architecture at Unite Us, he cut the implementation time for a national social-determinants-of-health platform from six months to under eight weeks while integrating with Cerner, Epic, and Allscripts across 14-plus states. His earliest executive work was just as hands-on. As Enterprise Architect at Blue Cross Blue Shield of Arizona, he migrated legacy claims and eligibility systems to Azure PaaS, cutting infrastructure cost by $2M and call-center volume by 48%. As Program Director at River Valley Behavioral Health, he took over a three-year stalled EHR rollout and delivered five go-lives in 18 months. As IT Manager at Luxottica, a $66B global retail brand, he delivered a $400M application and infrastructure migration in 14 months and cut projected data-center cost by $130M over five years, while running an offshore support center in China for 10,000 stores. He holds a Master of Science from Northern Kentucky University and a Bachelor of Arts from Thomas More College. He is known for cutting the forty-box architecture diagram down to the three things that matter, and for designing the human-in-the-loop and audit posture that keeps an AI program defensible when a regulator asks to see the math. If he cannot personally defend the output, he does not delegate the action.