Integrations Software Engineer

We're an AI-native engineering team: frontier models at high reasoning effort are the default tool for writing and reviewing code, and we're extending our own product into our own delivery loop wherever it fits. This is a time-sensitive engagement. We move on fast-follow cycles and need someone who can start immediately and deliver quickly — not a multi-month ramp-up. What you'll work on: - New adapters against our existing integration-layer pattern: each provider implements a common adapter interface that normalizes provider events into our engine's event vocabulary — no engine or workflow changes required per integration. - Jira & Linear as issue-tracker sources: inbound triggers, posting plans/results back onto the issue, and mapping each provider's event/webhook model onto our normalized types (we've already proven this pattern end-to-end). - Slack & Microsoft Teams as two-way chat surfaces: reading @mention-style triggers and approval replies out of a channel, and writing run progress, results, and elicitation prompts back into the same thread — the write-back path is what makes teams trust the agent, so it's not an afterthought. - Bitbucket as an additional source-control adapter, following the same shape as our GitHub/GitLab adapters: webhook verification, PR-event mapping, check/status write-back. - Credential handling for each new provider through our existing secret-storage stack (KMS-wrapped envelopes, per-provider connection metadata, rotation-aware decrypt) — extending it, not replacing it. Stack: Go, Postgres, webhook-driven event ingestion, OAuth/PAT auth flows, and hands-on familiarity with at least two of: Jira/Atlassian, Linear, Slack, Microsoft Teams, or Bitbucket APIs. Good fit if you: - Have built a real integration against at least one of these platforms (webhooks, auth, and write-back — not just a read-only API poller). - Work AI-natively yourself: you pair with frontier coding agents (Claude Code or equivalent) as your default way of writing code, and you're comfortable being reviewed by, and reviewing with, AI as well as humans. - Can implement to a defined adapter interface rather than freelancing your own architecture, and think about credential rotation and webhook-signature verification as first-class concerns. - Are available to start immediately and can commit to a fast delivery pace — this is not a slow-burn side engagement. Engagement: Contract, remote, immediate start required. [2 months / 20-40 hours per week]