Roman Swietlik

WORK HISTORY

EDUCATION

SKILLS

ABOUT ROMAN SWIETLIK

Fifteen years is long enough to have seen what actually makes products succeed. It\'s rarely the technology.At Arteria AI (Series B,$50M+), I partnered with data scientists and engineers to design a model-agnostic document intelligence platform, build LLM pipelines, and ship AI features that enterprise legal and financial teams actually rely on: 95%+ contract field extraction accuracy, processing time from 15+ business days down to 2–4. Real AI product work. A far cry from \"added a chatbot to the roadmap.\"Building my own software with Claude Code, Claude Desktop, and Codex is a daily practice, not a chat session. It exposes failure modes no PM spec sheet captures: models that answer confidently and incorrectly, accuracy that holds on clean data and falls apart on the real thing. That changes what I put in a PRD, what I push back on in an architecture review, and what I flag before QA discovers it six weeks later.20+ enterprise clients made one thing clear: the hardest part is never the technology. It\'s surfacing what users actually need before engineering starts, and keeping engineering, design, and business stakeholders pointed at the same goal all the way to delivery. Quest Diagnostics, TD Bank, Arrow Electronics, Canon, IKEA. MarTech ecosystems, enterprise SaaS platforms, native iOS/Android apps. Teams of 7 to 20, and I\'ve been the person at the center of all of it. I genuinely love it.Users don\'t care about your roadmap. They have real problems to solve. Solving those problems consistently is what drove 500K net new loyalty members for IKEA in under two months, a 46% engagement lift for Quest Diagnostics, and a 10% YoY increase in loyalty member share of sales.If you\'re building something real in enterprise B2B SaaS and need a PM who can hold their own in an architecture conversation and a boardroom in the same day, I\'d love to connect.