Perplexity AI Engineer / Search & Agent Systems Engineer
Open a video or resource link, watch it, then check its box to track your progress and unlock what comes next.
- Build grounded search experiences and understand Perplexity's agent-first application model
- Turn search and frontier-model capabilities into composable tools, context and durable workflows
- Engineer secure, evaluated and observable agents at search scale
Benchmark: Perplexity Agent Capabilities + Applied AI / search-platform engineering, paired with current Agent API, Search and Computer signals (current Sep 2026). Market refresh: Agent API and Search as Code; real-time web retrieval, URL/code/MCP/finance/people tools and citations; multi-model routing; Computer Skills, workflows and artifacts; long-running agents, memory and connectors; SPACE sandboxes; retrieval/ranking; trace replay, evals and production reliability in Python/Go/Rust/TypeScript/AWS.
Search Live Openings
We don’t have a direct job-board feed matched to this one yet, so here are real, live US job searches for “Perplexity AI Engineer / Search & Agent Systems Engineer” instead.
Build grounded search experiences and understand Perplexity's agent-first application model
Search API, citations + retrieval quality
Coverage: Search API / Agent API / Web Retrieval / URL Fetch / Citations / Ranking / Freshness / Filters / Structured Results
Agent API + multi-model orchestration foundations
Coverage: Agent API / Presets / Model Routing / Reasoning Effort / Search as Code / Tool Use / Python SDK / Responses Compatibility
Turn search and frontier-model capabilities into composable tools, context and durable workflows
Skills, MCP/connectors + long-running workflow composition
Coverage: Skills / Workflows / Artifacts / MCP / Connectors / APIs / Scheduled Tasks / Tool Discovery / Human Review
Enterprise retrieval + data-intensive agent products
Coverage: Deep Research / Data Warehouses / Data Map / SQL / RAG / Governed Context / Finance Search / Structured Artifacts
Engineer secure, evaluated and observable agents at search scale
Sandboxing, permissions + secure agent architecture
Coverage: SPACE / Ephemeral Sandboxes / Credential Isolation / OAuth / RBAC / Secure Connectors / Prompt Injection / Human Approval
Search-system performance, memory + eval-driven improvement
Coverage: Retrieval/Ranking / Memory / Agent Evals / Trace Replay / A-B Testing / Quality-Latency-Cost / Failure Analysis / User Signals
Role-Readiness Proof / Capstone
Build a production search-and-agent application on Perplexity. Use Agent API with an explicit quality/cost preset or model-routing choice, real-time web search and citations; add URL or code execution plus at least one MCP or enterprise connector; persist a bounded memory/context layer and compose a reusable Skill or workflow for a long-running task. Add one governed warehouse/private-data path with identity-aware access and structured artifacts; run code only in an isolated sandbox, keep credentials outside execution, and require approval for sensitive actions. Create offline and online evals for retrieval quality, citation correctness, tool selection, task completion, latency, cost and user satisfaction; capture traces and replay failures; test prompt injection, connector failure and stale data; then ship with monitoring, rollback and a measurable product-impact readout.
Proof Rubric
- Demonstrate grounded and cited answers
- retrieval and ranking judgment
- correct Agent API, preset and model choices
- robust tool, MCP and connector abstractions
- permission-aware enterprise context
- useful Skills/workflow composition
- safe sandbox and credential boundaries
- resilient partial-failure and retry behavior
- traceable long-running execution
- production-faithful evals and replay
- quantified quality, latency and cost
- effective prompt-injection defenses
- measurable task success/user value
- and evidence that evals, traces or feedback changed the agent