SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Higgsfield AI is building backend infrastructure for AI agents that execute multi-step creative tasks reliably in production. This role focuses on the execution layer: orchestrating agent workflows, managing tool integrations, enforcing permissions and resource limits, and ensuring recovery from failures.
You will own the full lifecycle of agent capabilities: understanding real user tasks, defining success metrics, building evaluation frameworks, shipping workflows, measuring impact, and iterating from production failures. Key responsibilities include:
- Agent Execution: Design and implement orchestration for multi-step tasks, tool calls, streaming updates, and cancellation logic.
- Durable Workflows: Build checkpoints, resumability, retries, and recovery mechanisms that prevent duplicate actions and ensure reliable completion.
- Tools & Integrations: Connect agents to generation services, editing tools, and external APIs through validated, secure interfaces.
- Execution Boundaries: Enforce user permissions, tenant isolation, resource limits, and approval workflows for high-impact actions.
- Evaluation & Observability: Trace failures, build regression test cases, and optimize for task completion, latency, and cost.
- Ownership: Move from unclear problem definition to production system, ship within days, measure real user impact, and improve based on failed runs.
This is a high-agency role. You'll choose priorities, remove blockers, cut scope to test ideas, and adapt based on evidence. The team values shipping fast, learning from production, and building systems others can operate and extend.
Requirements:
- Demonstrated ability to take a difficult, ambiguous problem from concept to production system, with clear explanation of your decisions and measurable results.
- Strong foundation in reliable agent execution: state machines, concurrency, queues, recovery patterns, model APIs, secure tool invocation, and repeatable evaluation.
- Self-directed problem-solving: you scope work, identify blockers, ask for help when needed, and persist through technical challenges.
- User-focused iteration: you investigate real user tasks and failed agent runs, use feedback to prioritize automation, and know when to stop or pivot.
- Collaborative mindset: you take feedback, change direction when evidence warrants, share knowledge, and enable others to operate and extend your work.
Nice-to-have experience: production agents or workflow engines, high-throughput systems with heavy traffic or long-running jobs, agent context management and memory, sandboxed execution and authorization, multimodal workflows (images, video, audio).
