AI Use Case Discovery Tool
Sellers had no reliable way to know which of Red Hat's marketed AI capabilities actually had real field traction. I built the framework and tool to close that gap.
01
The Problem
Red Hat's partners, Cisco, Dell, Lenovo, and others, market broad AI capabilities. But almost none of that maps cleanly to what's actually being sold, resourced, and repeated in the field.
Sellers were operating on marketing claims, not ground truth, which meant wasted pitches, misaligned assumptions about what Red Hat's portfolio could credibly support, and missed deals. My job was to figure out what was actually happening in the field and then decide what to build so sellers could act on it.
02
What I Ruled Out
My first instinct was to build a comprehensive database: every partner, every use case, every certification, fully cross-referenced. After interviewing PAMs and partner leads, I decided to modify that plan.
A perfect database nobody opens before a call is worthless. The real constraint wasn't information completeness, it was seller adoption in the flow of their work. That reframe changed the deliverable from a reference doc to a tool designed around how sellers actually research before a pitch. I also considered scoping the inventory broadly across all partner verticals from day one. Instead, I chose to stress-test the framework against a single regulated, high-stakes vertical first: fraud detection in financial services. The purpose was to validate the model held up before scaling it.
03
Discovery & Insights
Going Straight to the Field
Interviews with PAMs and partner leaders confirmed a real disconnect. I mapped it into a formal inventory: Partner Hierarchy, NVIDIA Focus Areas, Target Vertical, Certified Hardware, Deployment Architectures (RHAIE), and Pre/Post-Sale GTM strategy, turning scattered claims into something sellers could query.
Financial Services Deepdive
To stress-test the framework, I applied it to a regulated, high-stakes vertical: fraud detection. Tracing the path from partner hardware (Cisco, Lenovo, Dell AI Factory) through MLOps platforms (OpenShift AI, vLLM) to the customer-facing app confirmed the inventory held up against real deployment complexity.
An Unmaintainable Database is Worthless
The inventory only mattered if sellers actually used it, before a call or as active research, not buried in a shared drive. That reframed the goal: not a database, but a tool designed for seller adoption.
04
Design Process
The finished tool is proprietary to Red Hat and is not available for public display.
05
Reflections
Stepping outside my Computer Science & Design Comfort Zone.
Coming from a CS and Design background, I had to learn how enterprise go-to-market strategy actually works, how partners structure sales motions, how hardware and platform decisions ripple into deployment, and how sellers make decisions in real time. That shift shaped every choice in the project, including the decision to kill the database idea early. NotebookLM was also new territory; getting it to turn messy field notes into clean, structured data took real iteration on prompt design. I learned to treat prompting as its own design problem, not a technical shortcut. The biggest shift was learning to think in tradeoffs instead of features: every design decision here was really a business decision about where seller time and trust should go.