Building AI and data productsfrom messy problems.
I'm Shiven Arya, a University of Michigan graduate working across product, data, and AI. I'm drawn to messy systems: fragmented workflows, unclear operations, scattered data, and products that almost work but not quite. I've built marketplace dashboards, ML pipelines, Snowflake tooling, prototypes, and AI-powered systems that turn ambiguity into something usable.
100+
enterprise tables profiled
Millions
marketplace auctions analyzed
450+
clinical records modeled
Product + Data + AI
systems built
Selected Work
How I work
Find the real constraint
I start by separating symptoms from causes: is the issue data quality, workflow design, incentives, tooling, or user behavior?
Build enough to learn
I'm comfortable moving from research to prototypes to scrappy technical systems. I'd rather test something real than over-polish a deck.
Translate between teams
I can talk to users, write product specs, work through data, and communicate technical tradeoffs without hiding behind jargon.
Make complexity usable
My bias is simple in front, complex behind: clean interfaces, clear metrics, and systems that make the next decision easier.
Experience snapshot
AI Data Readiness
Built Snowflake-integrated profiling pipelines across 100+ enterprise tables and translated 14 data-quality dimensions into an AI readiness scoring system.
Marketplace Analytics
Built Tableau and SQL/Python analytics for auction health, surfacing marketplace KPIs like CTR, eCPM, win rate, and margin across millions of daily auctions.
Healthcare ML Systems
Built fraud-risk and outcome modeling workflows that turn provider and patient data into probabilities, clinical feature rankings, and plain-English outputs for non-technical stakeholders.
Notes on products, systems, and data
Let's talk.
I'm looking for full-time roles where product judgment and technical execution both matter. If you're building in AI, data tooling, marketplaces, healthcare, or workflow automation — I'd be interested in talking.