Joe Dymnioski
Project & account leadershipLead PM · Account DirectorWoven Health CollectiveAgentic engineerAI & Digital Strategy DeveloperCreator of MARIOWoven’s agentic harnessMLR · Veeva · oMAP · eMAP90% first-round approvalØ 9 yrs medical communicationsProject & account leadershipLead PM · Account DirectorWoven Health CollectiveAgentic engineerAI & Digital Strategy DeveloperCreator of MARIOWoven’s agentic harnessMLR · Veeva · oMAP · eMAP90% first-round approvalØ 9 yrs medical comms

Generative AI is easy to demo but hard to make dependable.

Dependable takes a system with people at the checkpoints.

In regulated work it has to be right every time.

Same brief, two outcomesPerson approvesPerson approvesA demo drifts off courseChecked at every gateProductionSame briefPersonapprovesPersonapprovesChecked atevery gateA demo driftsoff courseProduction

I built MARIO to be that system.

Woven Health Collective uses it every day, in regulated production.

See how MARIO works
OutputEmail, web, kiosk and toolsGate 2Expert review and approvalAgentsBuild, review and hardenGate 1Expert review and approvalPlanScoped, then held for reviewLesson library750+ reviewed lessonsStandards & SOPsWoven standards, 7 stacksInputThe brief and its documentsOutputGate 2AgentsGate 1PlanLesson libraryStandards & SOPsInput

On a real project, it went from kickoff to live in under two months.

A Phase 3 clinical-trial recruitment microsite: first client review on day 6, submitted to MLR on day 13.

LiveCalendar days from kickoff0714212835424956Day 6First client reviewDay 13Routed clean, submitted to MLRUnder two monthsLive, with analytics runningThe usual route: several months, just to reach MLRLiveCalendar days from kickoff0714212835424956Day 6First client reviewDay 13Routed clean, submitted to MLRUnder two monthsLive, with analytics runningThe usual route:several months, just to reach MLR

Two other systems bring the same discipline to editorial review and planning.

Multi-agent editorial screening

Specialist agents take the first editorial pass on regulated content, so editors start from a marked-up document instead of a blank review.

Projected 40%+ reduction in first-pass editorial effort

Complete. Built with Python, Claude Agent SDK, Custom PDF MCP server, Multi-agent orchestration.

Enterprise portfolio planning platform

It pulls projects straight from Workfront into Gantt views Woven's teams plan and manage from.

Project and portfolio Gantt views, resourcing, milestones and a planning mode, with client-facing views to share.

In development. Built with React, TypeScript, Vite, Express, Prisma, PostgreSQL, Socket.IO.

All three are built by someone who has sat through MLR.

Nine years running launch and program teams in agency medical communications, across oncology, cell and gene therapy, hematology, immunology, rare disease, neurology and vaccines. That includes the KYMRIAH launch and Pfizer’s COVID-19 sites for COMIRNATY and PAXLOVID.

ETHOS & AxiomProgram ManagerHavas LifeSenior Project ManagerVaniam GroupLead Project ManagerFCB Health & ProHealthAccount Director, IPG HealthWoven Health CollectiveAI & Digital Strategy DeveloperNowETHOS & AxiomProgram ManagerHavas LifeSenior Project ManagerVaniam GroupLead Project ManagerFCB Health & ProHealthAccount Director, IPG HealthWoven Health CollectiveAI & Digital Strategy Developer

I’ve led delivery teams and trained people beyond engineering to work with AI.

500+
people trained
100+
employees working from SOPs I wrote
20+
launch teams
One wayof workingProject managementAccountEditorialCreativeMedicalDigitalOne way of workingProject managementAccountEditorialCreativeMedicalDigital

Want to talk AI? Send me a note.