SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 188,500 - 282,700 / annual
Rubrik's Enterprise AI Incubation team is building agentic systems that reshape how enterprises operate. You will own production AI agents end to end—from understanding workflows and designing architecture through writing code, deploying, and running systems in production.
You'll work on real enterprise problems: designing multi-agent orchestrations that carry work end to end, such as agents that qualify and route deals, move orders through approval workflows under business rules, monitor customer health signals and recommend actions, or coordinate hire-to-retire processes across HR, IT, and management with shared state and clear handoffs.
Key responsibilities include:
- Lead product discovery with Product Managers and domain experts to map workflows, define problems, baselines, and success criteria for agentic solutions.
- Design, build, test, and deploy production agents and automations using LLM APIs, agent frameworks, and Rubrik's internal agent platform, integrated with business systems.
- Engineer agents for reliability: design tool flows and orchestrations, manage context and state, implement caching, retries, and failure handling.
- Build full-stack systems across backend services, APIs, data models, approval UIs, dashboards, and chat interfaces.
- Apply software engineering rigor to AI systems: treat prompts, skills, and configs as versioned artifacts with code review, tests, CI/CD, and staged rollouts.
- Implement evaluation, guardrails, and quality control: derive evaluations from workflow data, establish quality gates, integrate human-in-the-loop approvals, and monitor production runs.
- Use measured results to evolve or retire agents and turn successful systems into reusable patterns and libraries.
- Exercise judgment on AI fit: decide when a problem needs an agent versus simpler automation, and make tradeoffs around model choice, latency, cost, and autonomy using data.
You'll have explicit authority to stop or re-scope work that won't deliver impact and will thrive in ambiguous environments with a low-ego mindset, leveraging AI coding tools as collaborative partners.
REQUIREMENTS:
- Around 7+ years of professional software engineering experience in production systems.
- Around 2+ years of professional, hands-on experience building AI agents and agent-powered production applications with measurable outcomes.
- Strong full-stack engineering background: designing, building, and owning services end to end, including backend APIs and integrations, data models, and user-facing interfaces (approval UIs, dashboards, chat surfaces).
- System design skills: break down complex problems, design APIs and data models, reason about latency, throughput, and failure modes.
- Knowledge of algorithms, data structures, access control, and data provenance.
- Spec-driven development: turn vague requirements into clear problem statements, write design docs, use Spec → Plan → Task → Code → Converge flow.
- Execution discipline: version control, code review, automated tests, CI/CD, staged rollouts and rollbacks, logging, metrics, and tracing.
- Build agents in code using LLM APIs, SDKs, and libraries (not just no-code/low-code tools).
- Design robust agent harnesses and multi-agent orchestrations: model tools and resources, manage context and state, keep runs reliable under real-world data and load.
- Establish continuous evaluation and safety mechanisms: turn production failures into test cases, gate changes on eval suite results, enforce guardrails.
PREFERRED QUALIFICATIONS:
- Knowledge/context graphs: entities and relationships, GraphRAG, hybrid retrieval.
- Agent frameworks and runtimes: Google ADK, LangGraph, OpenAI style, cloud or container platforms.
- Backend: Python, FastAPI or similar stacks.
- Tools integration: MCP, Slack or Teams bots, REST APIs, webhooks.
- Data systems: Snowflake, Salesforce, graph stores.
- UX: conversational flows, chat surfaces, guided forms, simple dashboards for agent output.
