Staff Backend Engineer – Connector Platform & Tooling

Perplexity; San Francisco (CA); On-site; USD 220,000 - 405,000; Full time

Job summary

Perplexity is seeking a Connector Platform Engineer to build the data layer powering agents across hundreds of integrations. You will design and implement the connector runtime, host MCP servers, and CLI-backed tools behind a single agent-facing interface.

You will extend the semantic layer, define tool discovery, ensure robust agent loops with retries and observability, and work with Security on credential isolation.

Qualifications

• Experience designing and building backend systems that run in production.
• Strong system design skills with scalable architectures and API integrations.
• Proficiency in Python, Go, or Rust and multi-language environments.
• Hands-on cloud infra experience (AWS, Kubernetes).
• Knowledge of OAuth/authorization, API connectors, or LLM agent tooling.
• Security-minded with safe trade-offs; collaborative problem solver.

Responsibilities

• Own design and implementation of the connector runtime and hosted tools.
• Extend semantic layer with schemas and metadata.
• Design tool-discovery and tool-selection surfaces for agents.
• Ensure agent loops are robust with retries, observability, and idempotency.
• Define authentication and credential isolation patterns for connectors.
• Build onboarding path with fixtures and evaluation suites.
• Set SLAs, observability, and incident response for an always-on integration surface.
• Collaborate with product and AI teams on interfaces and patterns.

Skills

Backend design & architecture

System design

Python/Go/Rust

Security-minded development

OAuth & authorization

Collaborative engineering

Tools

AWS

Kubernetes

API gateways

MCP-server

Job description

Perplexity is seeking a Connector Platform Engineer to build the data layer powering agents across hundreds of integrations. You will design and implement the connector runtime, host MCP servers, and CLI-backed tools behind a single agent-facing interface.

You will extend the semantic layer, define tool discovery, ensure robust agent loops with retries and observability, and work with Security on credential isolation.