Platform Engineer

Take-home test project — 1 day, scope test here

It will be 4 week contract.

About the role

We are hiring one platform engineer to deliver nine components of a new AI agent discovery platform inside a 4-week MVP build. You will partner with one other engineer (research-heavy) on schema design and pipeline integration. Configuration-heavy work, fast execution, end-to-end ownership of every component you touch.

What you will build

Nine components across four weeks:

• Kubernetes test cluster (EKS or local kind) with an observability daemon (Pixie) deployed via Helm.
• Cloud inventory pipeline (CloudQuery) against a test AWS account — IAM roles, Bedrock inventory, CloudTrail filters.
• SaaS audit ingestion pipeline (Vector) — Okta System Log via OAuth, filtered for AI-service grants.
• Stream normalization layer (OpenTelemetry Collector) with three custom processors mapping inputs onto a canonical schema.
• Graph database (Apache AGE on PostgreSQL) with agent-graph schema and Cypher write path.
• Online machine learning baseline learner (River) wired into the streaming pipeline.
• Policy engine (OPA) with three Rego policy templates.
• Static analysis rule pack (Semgrep) for agent SDK imports, wired to a GitHub App.
• Read-only single-page demo UI on the agent graph (shadcn/ui + Tailwind).

Required experience

• Mid-to-senior platform or data engineer.
• Production experience with Kubernetes (Helm, EKS or equivalent) and AWS (IAM, CloudTrail, Aurora).
• Comfort with streaming and telemetry tooling: OpenTelemetry Collector (custom processors), Vector, or equivalents.
• PostgreSQL fluency. Graph database experience a plus (Apache AGE, Neo4j, or Memgraph).
• Python fluency for online-ML and processor work.
• Rule and policy authoring: OPA / Rego, Semgrep, or similar.
• Able to ship a small functional UI on a deadline (shadcn/ui + Tailwind or equivalent).

Key traits

• Fast execution. The build is 4 weeks. Buffer is 35%. Slippage compounds across engineer handoffs.
• End-to-end ownership. You own each of your 9 components from setup to integration to demo.
• Bias toward making OSS work as documented rather than rewriting it.
• Clean handoffs. Two named handoffs to the partner engineer fall on Day 6 and Day 9 and cannot slip.

Outcome

By the end of week 4: all nine components running and integrated with the partner engineer's work. The end-to-end demo scenario — a test agent making real API calls to an LLM provider and a database — produces a discovered agent record, anomaly findings, and a policy recommendation, viewable in the demo UI, in under five minutes.

Engagement

• 4-week contract, full-time.
• Remote-friendly.
• Hiring decision based on a 1-day take-home test project (provided after initial recruiter screen).

Hiring process

1. Recruiter screen — 20 minutes.
2. Take-home test project — 1 day, scope test here