Founder and principal
Stephen McDaniel leads YakData. His career has crossed product management, software development, data architecture, analytics, client delivery, development leadership, and production AI. That range matters when the expensive failure is rarely isolated to the model.
Selected operating background
Operator, not commentator
A consequential AI system crosses business value, data, architecture, software, evaluation, workflow, controls, economics, and operating ownership. YakData's advisory model is built around seeing those interactions rather than optimizing one component in isolation.
Product leadership at Brio, SAS, Yahoo, and Tableau included new products, established product lines, customer requirements, prioritization, tradeoffs, and responsibility for products generating substantial annual revenue.
Analytics leadership at Netflix and later work at Microsoft reinforced the same operating discipline YakData uses now: begin with the decision, establish the evidence, and work backward into data, analytics, and implementation.
Work has included Oracle consulting, regulated pharmaceutical data systems, data-warehouse capability at Opsware, worldwide data-ingestion architecture for U.S. Navy Cyber, server-side systems, scalability, security, and production data platforms.
For five years Stephen led a roughly 40-person SAS software organization spanning engineering, management, UX, testing, and documentation. The job was not to replace specialists. It was to keep the product, architecture, implementation, release, and customer result coherent.
Production evidence
In one production AI optimization, the system retained approximately 98% of the top model's measured performance while reducing model cost by 65%.
The decision was not “which model is smartest?” It was “what performance does the system actually require, what failures matter, and what economics survive production?” That is the level at which YakData works.
What this changes in an engagement
Cross-functional experience changes what gets noticed early, what gets challenged, and which risks are treated as first-class production concerns.
When the output is wrong, YakData also examines context, semantics, retrieval, workflow, system boundaries, acceptance logic, and human control.
Security, latency, cost, integration, maintainability, review burden, and failure recovery are treated as design inputs, not cleanup work after the demo.
The business consequence, system evidence, engineering tradeoff, and release decision are kept in one chain so the recommendation can survive scrutiny.
Trusted to teach experienced professionals
Teaching is useful evidence here because advisory depends on making difficult reasoning legible to executives, engineers, analysts, and operators who have to own the decision afterward.
INFORMS · American Marketing Association · TDWI
Princeton University · Brown University · University of Chicago · University of Washington, Seattle · North Carolina State University
Built, simplified, published
The books and teaching work are not the advisory offer. They are evidence of a repeated ability to find the governing structure inside complex technical work and communicate it clearly.
Stephen looks at what the business needs and works back to the technology. This insight is invaluable for improving ROI and driving quick wins from the analytical investment.
John LodmellCFO, Nordstrom Credit Services
The reason to hire YakData
It needs a defensible answer to a consequential system question, grounded in production evidence, tradeoffs, economics, and explicit ownership.