What the team does

Recart runs on ~40 micro-services across AWS, Kubernetes, and MongoDB. The Platform team makes sure that infrastructure is reliable, secure, and as invisible as possible to the engineers who depend on it. When a CircleCI build fails, when a dead-letter queue needs to be redriven, when an engineer needs temporary production access to fix a critical bug, Platform handles it.

But they're not just a reactive ops team. They build internal tooling that changes how the whole engineering org works. One of their biggest initiatives: a Windmill-based internal developer platform with a Slack bot interface, so engineers can request and receive temporary AWS IAM permissions or MongoDB access without ever leaving Slack. Approved, provisioned, and auto-revoked, all automated. The result is that the entire engineering team moves faster, with less bottleneck and more audibility.

They're also deep in the AI tooling wave: setting up Claude Code across repositories, generating claude.md files for each service, and integrating MCP servers so engineers can interact with Shortcut and other tools directly from their coding environment.

What they work on

• Cloud infrastructure - AWS resource management, IAM policy administration, SQS/SNS configuration, MongoDB Atlas clusters, SingleStore cluster scaling and performance tuning
• Kubernetes & GitOps - managing Kubernetes deployments via Flux (GitOps), handling cluster upgrades, scaling, and IOPS configuration
• CI/CD pipelines - CircleCI pipeline maintenance, build failure investigation, release process reliability across all repositories
• Internal developer platform - Windmill-based self-service tooling: automated IAM permission grants/revocations, MongoDB user creation, SQS dead-letter queue management, all via Slack bot commands
• Dependency management - Renovate bot configuration across all repositories, keeping the dependency graph healthy and PRs flowing
• Cloudflare - traffic management, security rules, redirect configuration
• AI developer tooling - Claude Code setup across the engineering org: claude.md generation, MCP integrations (Shortcut, Windmill), and building the workflows that let engineers use AI tools effectively in their local environments

Tech they work with

• Cloud: AWS (IAM, SQS, SNS, S3, EC2), Kubernetes, Flux (GitOps)
• Databases: MongoDB Atlas, SingleStore, Postgres
• CI/CD: CircleCI, GitHub, Renovate
• Internal tooling: Windmill, Windmill bot (Slack-native)
• Networking: Cloudflare
• AI tooling: Claude Code, MCP integrations

How our engineering teams work

On any given day, our engineers are fielding questions from CS, syncing with Product, and reviewing each other's scopes on Notion before a single line of code is written. We do standups, run design reviews, and write internal docs when something new gets built, so knowledge doesn't live only in Slack threads.

We move fast, but we're deliberate: new features go through feature flags, changes are scoped before they're picked up, and on-call rotations mean someone's always got the weekend covered.

What kind of problems they solve

"Every time an engineer needs temporary production access, it goes through a manual approval chain and takes half a day. Automate it."

"We need to dump the messagelogs MongoDB collection and restore it into SingleStore. Scale up the cluster, manage IOPS, execute the migration without downtime."

"Renovate stopped opening dependency PRs across multiple repos. Figure out why and fix it before the dependency graph drifts."

"Set up Claude Code for the entire engineering org. Generate claude.md files for 40+ services, wire in MCP integrations, make it work."

Who thrives here

You'll do well here if you find satisfaction in building systems that make other engineers faster, and don't need credit for it. The Platform team's best work is invisible: nobody notices the deployment pipeline, the access management bot, or the cluster configuration until it breaks. That's the goal.

Deep AWS and Kubernetes knowledge matters, as does a bias toward automation over manual processes. The Windmill bot is a good example of the team's philosophy: if you're doing something manually more than twice, build the tool that does it for you.

Interest in AI developer tooling is increasingly relevant. The team is actively expanding how engineers interact with Claude Code and MCP tooling across the stack.

Size & structure

2 engineers, working closely with the CTO.