Senior Software Development Engineer, Agent Workflows, Firefly Boards

Adobe Adobe · Enterprise · San Jose, CA

Senior Software Development Engineer to own the workflow layer for Firefly Boards' agentic capabilities. This role focuses on building and managing the agent workflows, including safety, quality, and cost controls, and ensuring survivability across interruptions. The primary focus is on the application layer that runs on top of the agent platform.

What you'd actually do

  1. Design and build the workflows that drive Boards' agentic product.
  2. Implement what the product needs, tune agent behavior for production use, and expand the agentic surface as new capabilities come online.
  3. Own the gates and guardrails on long-running tasks.
  4. Design and build the safety, quality, and cost controls that wrap agent operations: when an agent can act autonomously, when it needs to pause or escalate, how it recovers from partial failure, how budgets and limits are enforced.
  5. Design for survivability across interruption and resumption.

Skills

Required

  • TypeScript
  • Python
  • backend systems intuition
  • state management
  • long-running operations
  • recovery from partial failure
  • observability
  • writing clean, testable code
  • contributing to a codebase that other engineers extend
  • written and verbal communication
  • shaping technical decisions

Nice to have

  • workflow orchestration runtimes for tool-using LLM applications (LangGraph or comparable)
  • production LLM systems: evaluation, observability, latency and cost optimization at the workflow level
  • MCP (Model Context Protocol) and agent tool design patterns
  • production Kubernetes experience
  • real-time collaborative or multiplayer applications
  • using observability and analytics data to drive engineering decisions
  • AI-augmented engineering workflows
  • sophisticated orchestration of agents in the development process

What the JD emphasized

  • own the workflow layer
  • survive interruption
  • recover from failure
  • behave predictably in production
  • own the gates and guardrails
  • Most production agent systems fail here
  • Design for survivability across interruption and resumption
  • checkpointed, paused, resumed, and recovered cleanly

Other signals

  • agentic capabilities
  • multi-step operations
  • survive interruption
  • recover from failure
  • behave predictably in production
  • workflow layer
  • safety, quality, and cost controls
  • survivability across interruption and resumption
  • checkpointed, paused, resumed, and recovered cleanly