AI Platform Engineer
- TypeScript
- Python
- Go
- AI
#middle
#удаленка
Polygon Labs
AI Platform Engineer
Формат работы: удалённо (EMEA - Remote / USA - Remote)
Чем предстоит заниматься
-Move agentic workflows into production, taking prototypes to robust systems for real users and owners, with SLAs (e.g. spec-to-PR agents, AI code review, incident triage, KYB review) -Build and manage MCP servers, including system design, hosting, security, authentication (e.g. per-user OAuth), and keeping a clean connector inventory for Slack, GitHub, Jira, HubSpot, and internal services -Develop autonomous agents that act and plan across tools through trust stages: shadow, suggest, execute, and autonomous -Embed guardrails into the platform, providing each agent its own identity, scoped credentials, audit logs, CI checks, approval gates, allowlists, and least-privilege access -Build evaluation frameworks (test sets, regression suites, live scoring in production) so every agent releases with quality metrics tied to business outcomes -Instrument agents for observability: tracing, logging, latency, cost, drift detection, and daily dashboards for Operators -Track AI spend across providers (Anthropic, OpenAI, OpenRouter), setting budgets, alerts, and spend attribution to teams, agents, and workflows -Support Operators by developing features from their requests, writing documentation, running enablement sessions, and building self-service tools -Create skills and plugins centralizing company knowledge (processes, standards, security, brand), versioning and reviewing these to ensure consistent, repeatable agent output -Proactively find and automate manual, repetitive processes across the company, measuring outcomes and iterating rapidly
Наши пожелания к кандидатам
-5+ years as a software, platform, or infrastructure engineer, including experience prior to AI coding tools. Solid grasp of system design, data modeling, testing, debugging, and code review -6-12 months shipping LLM-based systems to production, with agentic workflows in use by real end users -Daily, hands-on experience with AI coding agents like Claude Code and Codex to write, test, and review production code -Experience building autonomous or multi-step agents with tool use (Claude Agent SDK, Anthropic/OpenAI APIs, or similar) -MCP server experience (OAuth flows, scoped tokens, user-level access control) -Identity and access experience: Okta (or similar IdP), SSO (SAML/OIDC), SCIM provisioning, group-based and device policies -Creating automated system identities (GitHub Apps, service accounts, API keys, secret managers like 1Password) -Designing guardrails: least-privilege, audit logs, CI checks, approval gates, allowlists, and policy-as-code -Creating evals for LLM systems: test sets, rubrics, CI regression, release gates, and live production scoring -Instrumenting production system observability: tracing, metrics, alerting, dashboards (OpenTelemetry, Grafana, Datadog, Langfuse, LangSmith, Braintrust) -Tracking AI spend across vendors: per-key limits, cost centers, budgets, actionable alerts -Strong in at least one of TypeScript, Python, Go; working knowledge of CI/CD (GitHub Actions/etc), cloud infrastructure -Track record of independently identifying and automating manual processes (two or three concrete examples) -Clear, concise written communication (specs, runbooks, RFCs)
Откликнуться
Подписаться на наши каналы / @best_itjob / @it_rab
What You'll Do
- Move agentic workflows into production, taking prototypes to robust systems for real users and owners, with SLAs (e.g. spec-to-PR agents, AI code review, incident triage, KYB review)
- Build and manage MCP servers, including system design, hosting, security, authentication (e.g. per-user OAuth), and keeping a clean connector inventory for Slack, GitHub, Jira, HubSpot, and internal services
- Develop autonomous agents that act and plan across tools through trust stages: shadow, suggest, execute, and autonomous
- Embed guardrails into the platform, providing each agent its own identity, scoped credentials, audit logs, CI checks, approval gates, allowlists, and least-privilege access
- Build evaluation frameworks (test sets, regression suites, live scoring in production) so every agent releases with quality metrics tied to business outcomes
- Instrument agents for observability: tracing, logging, latency, cost, drift detection, and daily dashboards for Operators
- Track AI spend across providers (Anthropic, OpenAI, OpenRouter), setting budgets, alerts, and spend attribution to teams, agents, and workflows
- Support Operators by developing features from their requests, writing documentation, running enablement sessions, and building self-service tools
openqareer не кадровое агентство: вакансия собрана с открытого источника, отклик подаётся на сайте работодателя.