I'm Dave Belding — a consulting manager at EY with nearly two decades of experience and close to nine years focused on AI/ML and data science. I architect the systems, not just the demos.
Background
I've spent nearly nine years embedding AI into large organizations — from Fortune 10 health insurers to global biopharma and athletic apparel brands. The work is part engineering, part translation, and a lot of knowing what not to build.
I lead data science and AI consulting engagements: architecting multi-agent orchestration platforms, semantic layers, and MCP gateways for Fortune 500 and Fortune 10 clients. My work sits at the intersection of enterprise constraints and frontier AI capabilities.
Outside of client work, I build agents, tools, and systems to stay sharp on the frontier. Recent projects span conversational AI, geospatial data pipelines, multi-agent frameworks, and MCP server development — all documented in the Projects section below.
I'm a certified Alpine Climbing Instructor and Ski Mountaineering Leader with The Mountaineers, and hold avalanche instructor certifications. The mountains are where I test my judgment when the stakes are genuinely high.
The gap between a polished AI demo and a production system that survives enterprise security, compliance, and change management is enormous. Most of my value — and most of my interest — lives in that gap.
Interactive
This conversational agent is trained on my LinkedIn profile, career history, and personal background. Ask it anything you'd ask me in a first conversation — it answers in-character.
Work
A selection of personal and course projects built to explore frontier techniques — and to stay grounded in what it actually takes to make them work.
A conversational agent that answers questions about my career, experience, and background using my LinkedIn profile and a personal context file as its knowledge base. Includes email capture for follow-up.
Live demo ↗An agentic recommender for hiking and ski mountaineering objectives. Integrates real-time weather forecasts, avalanche center bulletins, and GIS data to surface relevant destinations based on conditions and user goals.
Built and deployed Model Context Protocol servers as part of enterprise AI platform architecture. Includes examples of tool-augmented agents, structured output pipelines, and integration with external APIs via the MCP standard.
A series of multi-agent architectures from Ed Donner's Agentic AI Engineering course, including deep research tools, agent orchestration frameworks, and a capstone trading-floor simulation with specialized sub-agents.
A skill-based system that generates Gantt chart PowerPoint decks from a YAML data file, enabling agentic project planning artifacts that are both human-readable and machine-editable.
Additional projects in progress, including semantic layer architectures and enterprise agentic platform patterns. GitHub repos with architecture-diagram READMEs in development.
GitHub ↗Writing
Practitioner perspectives on enterprise AI — the systems, the trade-offs, and the gap between what gets demoed and what actually ships.
Writing in progress. Topics will cover enterprise AI implementation, agentic architectures, and what it actually takes to bring frontier AI into regulated industries.