Description
The Monetization team is building AI agents into their day-to-day workflows and extending that capability across the broader organization to enable sellers on pricing and packaging. You will be the internal steward of this agent workflow system, maintaining it, extending it, and working with the team to find where AI-assisted workflows create the most value.
This role sits at the intersection of engineering and analysis — requiring the business lens of a product manager and the technical depth to design, build, and evolve AI agent workflows. The ideal candidate thinks in systems, understands how teams operate, and can move fluidly between strategy and execution.
What you'll do:
Define the technical architecture for our agent portfolio (integration patterns, platform choices, guardrails)
Design and govern multi-agent orchestration patterns — how agents hand off tasks, share context, and operate within composite AI systems
Own the "connective tissue" layer: how agents, APIs, workflows, and enterprise platforms (Agentforce, Slack, Snowflake, Claude) are wired together and scaled reliably
Set standards for how agents are built, tested, and evaluated
Own agent performance post-deployment — monitoring, iteration, quality
Make platform calls (Agentforce vs. custom, Slack-native vs. standalone, build vs. buy)
Own agent identity and session design — author and maintain agent context files (identity files, session startup protocols, memory structures, and skill definitions) that govern how agents behave consistently across sessions, users, and evolving workflows
Manage system health and reliability — monitor, log, and handle errors; keep the agent infrastructure running as team dependency grows
Own the deployment lifecycle — manage configuration changes, deploy updates, and debug issues in the production environment without requiring outside engineering support
Experience:
Built or deployed LLM-based agentic workflows (not just chatbots)
Hands-on experience with agent orchestration frameworks (e.g., LangChain or equivalent) and multi-agent coordination patterns
Treats prompt engineering as foundational but not sufficient — one component within a broader orchestration architecture
Thinks in systems, not demos — cares about maintainability, evaluation, and scale
Translates fluently between strategy/business and engineering
Working Python proficiency and Linux comfort — can read the codebase, make configuration changes, and operate the infrastructure stack (environment config, systemd services) independently
LLM context file design as a craft — knows how to write durable agent context (identity, memory, startup protocols, skill definitions) that holds up across sessions — not one-off prompts, but role-level design
Self-sufficiency — able to deploy, debug, and iterate without waiting for outside engineering help; owns the full loop from design to production
Domain fluency in monetization workflows — You don't need to be a pricing expert, but you do need to quickly understand how the team thinks about deals, contracts, discounting, and revenue.
*LI-Y
For roles in San Francisco and Los Angeles: Pursuant to the San Francisco Fair Chance Ordinance and the Los Angeles Fair Chance Initiative for Hiring, Salesforce will consider for employment qualified applicants with arrest and conviction records.