I moved from solving customer problems on the front line to building the products that prevent them — now exploring how AI gets us there faster, without losing sight of what the data actually says.
Off the clock: I'm intuitive, a quick learner, and still a bit of an artist — I used to paint, and I've done portraits I'm genuinely proud of. These days most of my curiosity goes toward what's new in tech and AI, and figuring out which of it is actually useful.
I moved from solving customer problems on the front line to building the products that prevent them — now exploring how AI gets us there faster, without losing sight of what the data actually says.
Before product, I was on the front line — as a support and tech specialist, I heard customer problems first: the unmet needs, the friction, the things that quietly broke their workflow. I wasn't just logging tickets. I was troubleshooting complex integrations with engineering, and more often than not, proposing better ways to solve the underlying problem — not just the symptom. That instinct to fix the "why," not just the "what," is what pulled me toward product. When the opportunity came, I took it without a second thought.
Moving from support into product meant learning to see the other side of the idea — not just what customers needed, but how that need becomes something real. I learned agile and scrum, the stages an idea moves through from concept to build, and what a product manager is actually responsible for at each one: discovery, business value, MVP scope, prioritization, estimation — and just as often, knowing when to say no. The bigger shift was learning to hold two perspectives at once — the client's and the business's — and find the proposals that genuinely serve both, not just one at the other's expense.
Today, I'm focused on doing that faster and with more rigor using generative AI — building prototypes and proof-of-concepts myself, without waiting on dev cycles, to test whether an idea actually holds up before committing resources to it. What matters to me isn't using AI because it's AI — it's figuring out where it genuinely helps the process, and where it's just added cost and noise. Every call still comes back to the data, not to what the model thinks is a good idea.
Multiple entity-matching engines existed across different products, each matching people in its own way. I led the consolidation into a single engine that handles every data point across the business, with a roadmap to add more.
Defined the factors and data points used to build relationships between people, and designed how relationship strength is scored and labeled (Strong / Moderate / Weak). Also led a derived-network capability that surfaces warm paths clients didn't know they had.
Designed and built a set of interactive, AI-assisted prototypes to test product ideas before committing engineering time — best-match selection, a screening review workbench, relationship-path explainability, an AI-assisted relationship-discovery flow, and a client-level "opportunity graph." Used to pressure-test ideas with stakeholders and get faster, sharper feedback than a static spec ever could.
Ran a k-means segmentation on a 5,000-respondent survey to move from three qualitative personas to statistically validated segments, then picked the MVP segment using a weighted market-attractiveness score. Paired it with a four-metric launch dashboard built around a single north-star metric.
An AI tool that turns contracts and EULAs into plain-language summaries for small business owners who don't have in-house legal support. Based on a real problem, built around a fictional scenario for the case study.
Concepts I designed and built myself to test product direction with stakeholders and clients. Each one runs on sample data. Try any one live, or expand the screenshot.
Fixing and sharpening what clients use today.
Moving Altrata from raw data to insights.
If you're working through something tricky in product, data, or AI and want to think out loud, I'd genuinely like to hear from you. Like-minded people welcome, no agenda required.