An unreleased component page, shared with the Material Design hiring team. Enter the passcode to open the spec.
An independent page by Warren Velazquez, built with M3 Expressive tokens. Not affiliated with Google. Obviously.
Warren Velazquez. Two years embedded at Google as the agency, then four managing agencies from the client side. Program management for design systems, 13+ years, documented here in the format this team already reads.
Builds the system, gates the brand, plans the work.
Currently at Dow Jones: a design system that absorbs new event sites as the company defines them, the brand gate for product surfaces, and Relay, the AI intake he built. Previously ran android.com through a major OS launch.
Six teams with competing priorities, six spreadsheet formats, thousands of cells.
That is his intake at Dow Jones. The teams email spreadsheets; Relay, which he built, reads them into one prioritized, resourced plan. Pressure is an input, not a state of mind.
Android 12 launched on a public date that could not move.
Roadmap, creative, content, deployment and localization across android.com, tv.google and wearos.google.com. The date held and the team did not burn out, which is the part a Google director put on record.
Ten requirements from the posting. Ten receipts. Tap any row to jump to its proof.
| The role requires | On record | |
|---|---|---|
| 01 | Drive UX design development and sprints with 1P partners | Ran design and development sprints and roadmaps for android.com, tv.google and wearos.google.com, embedded at Google for two years. behavior/immovable-date |
| 02 | Manage third-party design agencies | Has sat in both chairs: was Google's agency at Huge, then managed Monaverse, Moving Brands and an Infosys team of 28+ engineers. behavior/both-chairs |
| 03 | Integrate UXR research into the process | A 3-month research phase before an 8-month design phase at APS. A year-long future-formats UX exploration with Google's Search UX Design Director. behavior/both-chairs |
| 04 | Execute within components with Eng and platform PgMs | Stewarded the APS design system into production with 28+ engineers. Currently migrating 8 core properties into one system at Dow Jones, 6 done so far, so new event sites are assembly, not invention. behavior/the-system |
| 05 | Manage narrative content programs | Owned content and localization for android.com, tv.google and wearos. Oversees content design and copy at Dow Jones today. release-history |
| 06 | Drive creative and icon-scale asset programs with cross-functional teams | The brand gate on product surfaces at Dow Jones, partnering with product design and commerce teams while overseeing design, content design and copy. Hundreds of assets through one stage-gated pipeline at GE. release-history |
| 07 | Own release schedules and milestones at high velocity | Android 12 on an immovable public date. About 8 events and 6 launches in the last 5 months, absorbed by the system rather than by heroics. behavior/immovable-date |
| 08 | Manage intake and prioritization with cross-functional partners | Six teams with competing priorities, six spreadsheet formats. He built Relay, the AI intake and planning layer that turns them into one resourced plan. behavior/intake |
| 09 | Ensure resourcing against key efforts | Owns the team's Asana capacity and resource system. "Meticulous capacity planning" per a Google director. behavior/cadence |
| 10 | Coordinate dependencies and communicate risks to leadership | One launch calendar across three Google properties. Review cadences at APS and GE built to surface misalignment early, not at launch. behavior/the-system |
Source: the Program Manager III, Material Design posting, Google Design Platform, Core. Every receipt links to a section of this page.
Dow Jones defines new events every other week. Warren leads the design system built to absorb that: one component set across 10+ event and conference sites for WSJ, Barron's and MarketWatch, so every new site is assembly instead of invention. Engineering time drops about 80 percent, and stakeholder review rounds and repeated design decisions drop with it, because the system already made them.
6 of 8 core properties migrated. The two outlined tiles are not animated because they are not done. This page does not round up.
On the 80 percent: that is engineering time to stand up a new event site, measured against the pre-system baseline, and it keeps its "about" because it is an estimate across sites rather than a single number off a dashboard.
Android 12 launched on android.com on a public date. Warren ran roadmap, creative, content, deployment and localization across android.com, tv.google and wearos.google.com, one calendar over three properties with shared components and independent release dates.
Drag the slider. The lanes compress and reflow around a launch line that is the only element on this page with no animation. The lane widths are schematic, not measured; the ordering and the overlaps are how the program actually ran.
Warren is the brand gate on product surfaces at Dow Jones, which means six teams send him their plans and requests in six spreadsheet formats. So he built Relay. The hard part was never sorting things into three buckets. It was reading intent out of prose, reconciling six vocabularies into one schema, and chasing the fields nobody filled in. Step through it below.
Subscription strategy, product design, product management, performance marketing, CRM and event marketing each send their own file. Asking for one template has never worked, so Relay meets them where they are.
Six teams, thousands of cells. The count is the reason this is a system and not an afternoon.
Requests do not arrive as fields. They arrive as sentences in a cell, half of them written in a hurry. This is the part a formula cannot do, and it is where the AI actually earns its place in the pipeline.
"Q3 paywall test, need design + copy, maybe eng? mid Aug, check with Sarah"
Everything on the right of this section is derived from a cell like that one. Nothing is re-typed by hand.
Three teams describe the same week three ways and the same surface three ways. Everything downstream depends on collapsing those into canonical values, because a plan built on six vocabularies is six plans.
The third row is the interesting one. Relay does not guess a missing field. It flags it and moves to stage 4.
This is the stage Warren cares most about. A tool that hands you a list of gaps has moved the work, not removed it. Relay drafts the follow-up, sends it, holds the request open, and fills the field when the answer comes back.
"Quick one on the Q3 paywall test: who is the owner on your side? Everything else is in."
Nobody was chased by hand. The null from stage 3 closes itself.
Every item carries a reason and a capacity cost, so the ranking is arguable rather than oracular. Items that are low confidence or over capacity do not get sorted quietly; they land in a tray that needs a person.
Where the human stays. Relay does the reading and the arithmetic. Priority and resourcing are a judgment call and stay a judgment call.
Planning finishes in Relay. Only then does the structured output cross into Asana through the API, which is also the reason the plan and the tracker cannot drift: one of them writes the other.
task Q3 paywall test project brand.q3 due_on 2026-08-15 surface wsj.subscription effort design 3d, copy 1d source subscription strategy
Warren also owns the team's Asana implementation, which is why the field names on the left already exist on the right.
Warren is not an engineer. He directs AI, and the point is not that he writes code. The point is that he can take a fuzzy operational problem and return a working system without waiting for headcount. Four of them, in four unrelated domains, which means the transferable thing is the method rather than the domain.
Six teams send quarterly plans in six spreadsheet formats, and reconciling them by hand ate the first two weeks of every quarter.
The AI-hard part. Reading intent out of prose, then collapsing six vocabularies into one schema without inventing the fields nobody filled in.
Where the human stays. Priority and resourcing. Relay proposes a ranking with reasons; Warren decides and every override is recorded.
The resourcing meeting lived in three places at once: the call, someone's notes app, and a board that got updated later or not at all.
The AI-hard part. Turning live speech into structured resourcing proposals, each one carrying the sentence that triggered it, so a suggestion can be audited rather than trusted.
Where the human stays. Propose-only, by architecture. Nothing lands on the board until somebody clicks Approve. Speech to text runs on-device so audio never leaves the room, and teams bring their own Anthropic key, so the analysis routes through their account and they see the bill.
vizor.workHiring conversations, vendor scoping and partnership talks all start before anyone has signed anything, so the commitments made in them are unrecorded.
The AI-hard part. Getting to a usable agreement in about sixty seconds, then keeping a record of every commitment made under it.
Where the human stays. Both parties agree before the conversation, not after. The counterparty graph is a memory aid, not a decision maker.
pacto.legalConfidential files get shared under an NDA that is signed separately, late, or never, and nobody can prove afterwards who read which page.
The AI-hard part. Tailoring an NDA to the specific share, and holding a tamper-evident ledger of every view, every page and every minute.
Where the human stays. The recipient signs before page one unlocks. Nothing about the gate is inferred.
looklock.ai"It turned our weekly planning meeting from a doc-readout into a real conversation, and now the conversation updates the board itself."
Why this matters for a program role and not just for a resume: every agency the Material team hires is now quoting AI-assisted timelines. Someone who has built these systems can tell a real efficiency claim from a markup, and can write a scope that does not get gamed. He builds these to make the program run, not to leave program management.
This role manages third-party agencies for the Material team. Warren has done that job, and he has also been that agency, inside Google, for two years.
Huge, Android pod, embedded at Google
Two years as Google's agency-side senior PM. Owned roadmap, creative, content, deployment and localization for android.com, tv.google and wearos.google.com through the Android 12 launch.
Google Search Ads futures
Led 4 to 6 designers for a year-long exploration of future ad formats, partnering directly with Google's Search UX Design Director. Influence with no formal authority, sustained for a year.
What that teaches
Exactly what a 1P partner needs from an agency brief, and exactly where agency work drifts when nobody manages the seam. He knows the failure modes from inside them.
Infosys, 28+ engineers
Stewarded the APS design system into production with a 28+ engineer vendor team, with a review cadence built to catch drift at implementation, not at QA.
Monaverse, end to end
Owned the vendor relationship for The Row at Everyrealm through implementation, platform testing and a single coordinated launch, plus external motion, 3D and video vendors.
Moving Brands and beyond
Agency-side engagements for NYU Langone's design system audit. He knows how agencies are scoped, priced and judged, because he has written both sides of that paperwork.
"Meticulous capacity planning to hit targets across every workstream without burnout. Exceptional communication and problem-solving."
"A team player and problem solver on a complex, multi-part, year-long program. Strong client relationships throughout."
"A great PM. Diligent on the financial, creative and timing details, and a strong voice on UX and design."
The last five months at Dow Jones: about 8 events and 6 launches, every one a fixed-date public go-live. The cadence holds because the system does the repetitive deciding and Relay does the digesting, not because anyone is sprinting.
The line is the pill count above it, nothing more: two, three, four, three, two. May carried four dated go-lives and was the month to watch. Capacity was planned in Asana with Relay feeding the plan, which is the whole reason a four-launch month was a peak rather than an incident.
Oldest first, against changelog convention, because the arc matters. It starts with Google and it keeps returning there.
First Google work: the Google for Work website, delivered on a tight deadline with a distributed team across time zones. Also Viacom's intranet portal and Novartis patient-app explorations.
The Petfinder rebrand and redesign through its launch campaign, one coordinated go-to-market. Managed the Purina Pro Plan Veterinary Diets platform used by vets worldwide.
The APS design system: a 3-month research phase, an 8-month design phase, then stewardship into production with Infosys and 28+ engineers. A review cadence that brought skeptical utility executives into the process instead of around it. Also Vodafone, SAP Ariba, Microsoft and Merck.
Two years embedded at Google. Roadmap, creative, content, deployment and localization for android.com, tv.google and wearos.google.com, including the Android 12 launch on an immovable public date. Separately led 4 to 6 designers on a year-long Google Search Ads future-formats exploration with Google's Search UX Design Director.
Built the creative operations function from zero at a $50M+ a16z-backed startup. Hired 3D artists, UI/UX, motion and visual designers. Shipped across 12 metaverse platforms. Owned the Monaverse vendor relationship end to end.
GE Healthcare's Better Health Study: hundreds of media assets through one stage-gated pipeline with executive readouts. NYU Langone design system audit with Moving Brands. A pharma site delivered inside merger-grade legal review. On the side, shipped three AI products solo: PACTO, Vizor and LookLock. Relay came later, at Dow Jones.
Brand steward across WSJ, Barron's and MarketWatch product touchpoints. Leads the design system above, built Relay, owns the team's Asana implementation, and runs the fixed-date launch calendar you just scrolled past.
Everything above this line is retrospective. This part is not, and it is offered as a starting position rather than a plan, because the real one gets written after the first month of listening.
The system is documented; the seam is not. Where do agency deliverables actually enter Material, who accepts them, and at which step does a component brief stop being legible to the person building it. Same question for the 1P side: what does a surface team send when they need something, and how many places does that request live before it becomes work.
Output: one written map of every intake path and every handoff, with the drop points named.
Intake is where a platform team quietly loses its quarter, and it is the thing he has built twice. One request shape, one queue, one place where priority is decided and recorded with a reason. Not a tool first. A schema first, then whatever tool the team already has.
Output: a single intake path in the team's existing tooling, and a prioritization record that survives the meeting it was decided in.
Components, guidance and the agency work that feeds both, against the platform dates that do not move. This is the Android 12 problem again at a different scale: shared components, independent release dates, one calendar. He has run exactly that, for two years, on Google properties.
Output: one calendar the whole team reads, and a standing risk review that surfaces misalignment early rather than at launch.
What he would need: an honest hour with whoever currently absorbs the intake, and permission to change the shape of a request before changing anybody's roadmap.
Material's own docs are honest about parity gaps. This page follows the convention, because a spec that only lists what works is a brochure.
Warren has never carried a Google badge. Two years embedded at Google through Huge is proximity, not membership, and the difference is real.
Relay has exactly one user and has not been hardened for a team. It is a working system, not a shipped product, and he would rather say so than let the diagram above imply otherwise.
He has seen Material from the surfaces that consume it, at Google scale, on a fixed public date. That is the seam this role manages, viewed from the side that feels it first. He has also spent 13+ years being the other end of this role's phone calls, and the fastest way to manage agencies well is to have been one. The rest is onboarding.
Warren answers his own intake queue. New York based.