Principal AI Systems Architect – Agent Orchestration

Leadout Capital; Massachusetts; On-site; USD 180,000 - 270,000; Full time

Job summary

Salesforce AI Labs is seeking a Principal Engineer to drive architecture for a production-ready agent system that learns from demonstrations and blends AI reasoning with deterministic execution. You will work directly with customers, shaping the product and guiding other engineers from design through operation.

You will lead architectural decisions, evaluate model choices, and mentor teammates while ensuring reliability, security, and maintainability of critical capabilities across agent and

Qualifications

• Led the architecture and delivery of complex backend systems across deployment, failures, and improvements.
• Hands-on experience shipping and operating LLM-powered systems and analyzing failures.
• Proficient with Python, Go, Java, or JavaScript/TypeScript.

Responsibilities

• Design and build systems that learn from demonstrations and combine AI reasoning with deterministic execution to complete workflows.
• Develop recovery and verification capabilities so agents adapt to changes, confirm outcomes, and indicate when human intervention is needed.
• Build backend services with robust workflow state management, failure handling, and secure access to customer environments.
• Create evaluations and diagnostics that measure task success, expose regressions, and make failures understandable.

Skills

Backend architecture

LLM-powered systems

Python/Go/Java/TypeScript

Engineering leadership

Customer engagement

Job description

Salesforce AI Labs is seeking a Principal Engineer to drive architecture for a production-ready agent system that learns from demonstrations and blends AI reasoning with deterministic execution. You will work directly with customers, shaping the product and guiding other engineers from design through operation.

You will lead architectural decisions, evaluate model choices, and mentor teammates while ensuring reliability, security, and maintainability of critical capabilities across agent and