How a Forward Deployed Engineer Helps Pet4Me Keep Shipping with Claude.

Pet4Me’s own promise is to make pet connections simpler and safer, without subscriptions, fees or gated messaging. People list pets for adoption and rehoming, report lost pets, announce studs and offer local services, and they coordinate through in-app messaging without sharing personal contact details upfront. The platform is monitored, listings are validated automatically before publication, and reporting takes one tap.

The first part of the story, how unicrew researched, designed and built Pet4Me, covers its journey from concept to a live product at pet4me.org. This part covers the work since then.

After the build phase, Pet4Me needed a different kind of team: one that could complete the platform, harden trust and safety and keep shipping improvements, with design, QA and project management still covered.

By then Pet4Me had a complete design system, a working platform on Laravel and Vue/Nuxt, and its four core feature layers: listings, communication, trust and safety, and pet profiles. A build team and a continuous-delivery team are not the same shape. Two engineers, a QA engineer, a designer and a project manager are sized for building the foundations in parallel. Finishing the platform and shipping new features on top of it goes fastest with someone who holds the full context and ships without hand-offs.

So unicrew proposed a Forward Deployed Engineer: one senior engineer from the build team, embedded in Pet4Me’s delivery, owning it end to end and working directly with the founder, with Claude covering the capacity of design, QA and project management. The other four members of the build team moved on to other unicrew projects. It works here for two reasons: the engineer is senior and carried the full context from the build, and the founder is closely involved, knows exactly what he wants and needed a fast pace, which one engineer working directly with him delivers better than a team with hand-offs.

A Forward Deployed Engineer powered by Claude owns the whole loop, from the code to the conversation with the founder, and Claude is what makes that loop fit one person’s week. The founder brings the requirements. The engineer turns them into a PRD, designs the screens, implements the feature, writes the tests and the documentation, and publishes it to staging, where the founder verifies it before it goes live. There is no hand-off to QA, no design ticket waiting for a mockup, no PM writing the status update. Each of those is a step the engineer runs with Claude:

• Engineering.

  Claude Code is the primary development environment. Features are implemented against the existing codebase conventions, tests are written in the same pass, and a second Claude Code pass refactors and optimizes before anything ships.

• Design.

  New screens and components start in Claude Design, fed with the UI kit from the build phase (typography, color tokens, spacing and component states). Claude Design produces layouts that match the established system, Claude Code turns them into production components, and the result is consistent with the screens the original designer produced.

• Quality.

  The test suites written alongside each feature are the automated QA function. Claude Code writes them, the engineer reviews them, and CI runs them. Acceptance happens on staging: the founder verifies every feature there before it is released.

• Project management.

  PRDs, task breakdown, documentation, release notes and client updates are drafted with Claude.ai and Claude Cowork and finalized by the engineer. The founder receives the same cadence of communication he had with a dedicated PM, now straight from the engineer who builds the product.

What keeps this accountable is simple: the engineer reviews what Claude produces and decides what ships, and the founder verifies every feature on staging before release. Claude is used to build Pet4Me: it works on the code, designs and documents, but never touches real user data. The product’s knowledge does not live only with the engineer: it sits in the test suite, the documentation and release notes, and the UI kit.

Messaging, security-reviewed with Claude Code

Messaging is where a marketplace earns trust, and Pet4Me’s lets people coordinate without sharing personal contact details upfront. The engineer hardened the messaging layer and ran a security review with Claude Code on every feature before it shipped, working through what an attacker would try and how the code holds up.

The inbox groups conversations by listing rather than by person, so every thread is anchored to what it is about: an adoption listing, a stud announcement, a lost pet report, a grooming service. Unread counts, delivery status and search across conversations are all there.

A scam warning inside the conversation

Pet scams follow a small number of well-known scripts. In a typical one the seller cannot meet in person, a third party handles delivery, and the buyer is asked to pay fees upfront before the animal ships. Nobody meets the pet. The money leaves. The pet never existed.

Pet4Me recognizes these scripts and attaches a warning directly under the message, in the thread, at the moment the user reads it. The platform tells them before they reply. The engineer built the detector with Claude Code, using Claude to research the well-known scam scripts it needed to recognize. The detector runs inside Pet4Me’s own platform: Claude helped build it and never sees users’ messages or any other real user data.

Listings screened before they go live

Every listing on Pet4Me, whether adoption, rehoming, stud, lost and found or a service, goes through automatic validation before it goes live. Every uploaded photo is screened for sexual content, drugs, violence, smoking and other categories that have no place on a pet marketplace, and every free-text field for abusive language, threats, violence and other prohibited content. A listing that fails is blocked with a clear message, and the user sees which field failed and why.

The aim is to block prohibited content before publication rather than remove it after a complaint, and user reports remain a second line of review. The first version of these filters shipped with the build, and the engineer extended and hardened them with Claude Code.

The promise cards, designed with Claude Design

Pet4Me’s homepage makes four promises: it is free to use, messaging stays inside the platform, suspicious behavior is detected early, and unsafe content is kept off the platform. The scam warning and listing validation are part of the last two. The cards that present them, with their numbered layout, outlined display numerals and alternating teal and amber panels, were designed in Claude Design from the build-phase UI kit and built with Claude Code, as were the messaging screens above.

SEO for a marketplace that lives on discovery

A free marketplace lives or dies on organic discovery, and the build chose Nuxt in part for its SEO capabilities. With Claude as a reviewer and reference, the engineer applied a full pass of SEO best practices across the platform, from page structure and metadata on listing and category pages to structured data, sitemaps and image optimization. Claude brought the checklist and the reasoning, and the engineer decided what fit Pet4Me and shipped it.

Many products reach this point: a finished design system and a working platform, with a roadmap of features still to build. That is where a Forward Deployed Engineer earns its place: one senior engineer who, working with Claude, runs engineering, testing, design and project management. It works when the foundations are in place and the product owner is closely involved and decides quickly, and then it is a fraction of the cost of carrying a full build team through the rest of the roadmap.

Pet4Me’s switch rested on an engineer who already knew the product. Two situations differ from Pet4Me’s. If another team built your product, the first step is a written assessment of what you already run, the way our software maintenance and support engagements start on systems we did not build. A greenfield product with no design system or architecture yet is a different job, and it needs a team built for parallel work.

Most engagements start within two to four weeks, and there is no minimum engagement period. What changes about roles and code review once AI sits inside a team is the subject of our guide to building AI-native development teams. For a different kind of Claude Code work on a live platform, see our CRM API migration for WaiverKing.