I’m a senior product engineer building AI-enabled web products—and a
technical lead who stays close to the code. My 10+ years of shipped work
spans public products, internal review workflows, and university platforms.
Public MetaSculpt interface. I co-founded the product and owned AI workflow design, crawler and reporting logic, documentation, and code review with a co-founder.Krystian Flores · Tucson, Arizona
10+ yearsshipping production web systems
Public AI productco-founded and shipped
< 1 minutefor clear logo-review cases, down from about 10 minutes
Selected work
Shipped work and the decisions behind it.
Start with MetaSculpt, the internal logo-review workflow, and public
university frontend work. Project status is shown on every case study.
Public MetaSculpt interface. I co-founded the product and owned AI workflow design, crawler and reporting logic, documentation, and code review with a co-founder.
Co-founded and shipped a public product combining prompt tracking, citation capture, crawl analysis, weekly reports, and Schema JSON generation and validation.
Led product architecture, AI workflow design, frontend implementation, crawler and report logic, documentation, and code review in a two-person company.
Designed the Crawl → Generate → Validate → Fix flow and kept crawler and generator responsibilities separate.
Internal logo-review workflow I designed. Deterministic checks and retrieved
policy context support the reviewer; ambiguous cases stay in human review.
Desktop and mobile admissions pages. I owned development between the shared university header and footer and mapped Drupal paragraph types to reusable Gatsby components.
Curious about the system. Practical about the outcome.
I’m happiest untangling a complicated workflow, making the interface feel
obvious, and leaving behind a system other people can keep using.
Start with the messy workflow
I look for the handoff, edge case, or editor constraint that decides whether a product works outside the happy path.
Make the system inspectable
Whether the output comes from a CMS or a model, people should be able to understand its state, review it, and decide what happens next.
Leave a reusable path
The strongest implementation gives the next developer or content editor a pattern they can safely use without me in the room.
After-hours build
Yes, I turned March Madness into a data product.
The NCAA 2026 Bracket Model started as an office-pool question and became
an end-to-end lab: public data collection, feature engineering, model
evaluation, Monte Carlo simulation, and a playful interactive interface.