Senior Security Engineer for Multi-Agent AI Observability & Tracing

Paid Technical Trial — Multi-Agent AI Observability & Causal Tracing We are evaluating senior engineers for a larger proof-of-concept project involving observability and security analysis across autonomous AI agents. This is a small, paid technical trial intended to assess architectural judgment and implementation quality before awarding the larger engagement. The trial should take approximately 6–10 hours. Objective: Build a minimal two to three-agent workflow in which one agent delegates work to another. Instrument the workflow so that we can reconstruct the causal chain across the agents. The key question is: Can we determine not only that multiple events happened near each other in time, but that one agent action actually caused or delegated the next? Trial Scope: Please build a small working example that includes: 2–3 autonomous agents; unique identity for each agent; one shared task/session identifier; at least one agent-to-agent delegation; at least one tool or resource interaction; event logging with timestamps; parent/child event relationships; trace or correlation IDs; a simple graph or structured output showing the full causal chain. Example: User Task ↓ Agent A ↓ delegates Agent B ↓ invokes Tool / Resource ↓ returns Agent B ↓ responds Agent A The implementation does not need a polished interface. Technical Expectations You may use: Python; OpenTelemetry; LangGraph, LangChain, OpenAI Agents, AutoGen, or another agent framework; PostgreSQL, JSON, or another lightweight event store; NetworkX, Neo4j, or another simple graph representation. Use whatever stack you believe is appropriate. We care more about clear causality and architecture than framework choice. Deliverables Please provide: Working source code. Setup / run instructions. Example telemetry output. A simple visualization or structured representation of the activity graph. A short explanation of: how causal relationships are preserved; what limitations exist; how the design would need to change for a larger multi-agent environment. Evaluation Criteria We will evaluate: clarity of the event model; preservation of causal lineage; quality of instrumentation; simplicity of architecture; code quality; understanding of distributed tracing; ability to explain tradeoffs; ability to distinguish causation from simple temporal correlation. Proposal Requirement In your application, please briefly answer: If five agents each perform authorized actions, what information would you need to capture to determine whether those actions were part of one coordinated workflow rather than five unrelated events? Please also include: your most relevant experience with distributed systems or observability; experience with OpenTelemetry or tracing; experience with AI agent frameworks; whether you personally will complete the work. Project Details Budget: $800 fixed price Expected effort: 6–10 hours Expected turnaround: 3–5 days after start This trial may lead directly to a larger 4–5 week paid engagement for the strongest candidate. All code and work product created for the trial will be owned by the project.