Stable node URL

Evals Engineer

The discipline role that keeps agent quality measurable and honest.

hybrid

pilot

hybrid

Senior

Aliases

Sources

Neighbors

What this page is for

A stable inspection page for one graph node.

• The path identifies this node directly.

• Compare state stays in the query string.

• Add interest saves this node on your profile for a later aligned path.

Compare with a neighbor

Advanced

Personality-aligned path

Use saved Role Lattice interests plus your Workplace Archetype Quiz result.

Node profile

Humans set thresholds. Agents draft and classify tests, but humans own the final quality line.

Role frame

Design reliable evaluation systems that reveal whether agentic behavior is good enough.

Obligations

Measure quality

Detect regressions

Make failure visible

Tasks

Build eval harnesses

Label edge cases

Track regressions

Compare model outputs

Skills

Testing

Experiment design

Metric definition

Data analysis

Tools

Eval dashboards

Test runners

Prompt logs

Annotation tools

Prerequisites

Engineering discipline

Product judgment

Agent eligible

Draft test cases

Cluster failure patterns

Summarize eval runs

Human gated

Pass/fail criteria

Quality threshold changes

Release sign-off

Comparison

Evals Engineer vs AgentOps Engineer

The control plane role for safely running agentic systems in production.

Kind

hybrid vs hybrid

Seniority

Senior vs Senior

Demand

256 vs 518

Space

pilot-space:agentops → pilot-space:agentops

Neighbors in common

Shared tools

Compare is meant to answer one question: what changes when I move from one node to the other?

Evals Engineer only skills

Testing

Experiment design

Metric definition

Data analysis

AgentOps Engineer only skills

Automation

Evaluation design

Policy framing

Incident response

Neighbor nodes

Connected nodes

Use these to jump to a related node or compare against the current one.

AgentOps Engineer

related · hybrid · pilot-space:agentops

The control plane role for safely running agentic systems in production.

Applied Scientist

related · human · pilot-space:data

Scientific rigor aimed at real product and operational improvements.

Invoice Reconciliation Agent

related · agent · pilot-space:agentops

A first-class non-human node that keeps repetitive reconciliation work structured and auditable.

Stable lookup

The explorer now resolves search hits to this node page. The compare query can be swapped without changing the canonical path.