SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Certivity builds an automated pipeline that collects regulatory updates from 80+ sources, processes them through extraction, embedding, search, classification, consolidation and translation. As the company expands to new regions, maintaining production reliability has become critical.
You will own platform reliability and observability for this data pipeline, working alongside a core developer. This is not a ticket-driven ops role—you'll write production Python from day one and take ownership of system health end-to-end.
Key responsibilities:
• Reduce alert noise by distinguishing transient failures (network blips, timeouts, self-healing errors) from real issues, and build alerting the team trusts.
• Design and build agentic incident response systems: agents that gather Sentry issues, logs, metrics and recent deploys, classify failures, apply known fixes or open draft PRs, and brief the right developer with root cause and suggested remediation.
• Monitor data quality and pipeline health, not just infrastructure. Track freshness and completeness metrics (e.g., "every source checked on time," "every document produced text and embeddings").
• Implement self-healing infrastructure: retries with backoff, idempotent and resumable jobs, dead-letter handling, and clear escalation when automation reaches its limits.
• Ensure agent safety through scoped permissions, audit trails, human approval for risky actions, and tracking of agent accuracy.
• Continuously audit infrastructure for efficiency gains and cost reduction.
• Catch memory, timeout and cost problems across Cloud Run before they become silent OOM kills or surprise bills.
• Add structured logging, metrics, tracing and sensible Sentry grouping, with infrastructure as code and CI/CD.
Tech stack: Python, Docker, GCP (Cloud Run, Cloud Logging, Cloud Monitoring), Sentry, MongoDB, Azure Blob Storage, LLM APIs. Some Go and React.
Requirements:
• 5+ years in software engineering, SRE or platform roles, with real production Python experience.
• Proven track record of turning an ignored alert channel into one the team acts on.
• Hands-on experience building with LLMs or agents in production, with judgment to know when a plain if-statement is better.
• Solid experience with GCP (or similar cloud platforms), containers and CI/CD.
• Strong grasp of distributed-system failure modes: retries, idempotency, partial failure and backpressure.
• Pragmatism and clear communication in a small team.
