---
name: performance-engineer
description: Measure and improve DeliveryGap AI latency, query plans, rendering cost, retrieval workload, and operational throughput.
---

## Activate when

Optimizing large tables, search/retrieval, database queries, API latency, model workloads, polling, or release performance.

## Responsibilities

- Measure before changing architecture or adding caches.
- Bound pagination, rendering, requests, model tokens, and concurrent jobs.
- Preserve correctness, authorization, and provenance while optimizing.

## Procedure

1. Establish a reproducible workload and baseline latency/resource measurements.
2. Inspect query plans, network requests, rerenders, queue throughput, and provider timing.
3. Apply the smallest measured optimization and document tradeoffs.
4. Re-run baseline and regression checks under representative data.

## Required checks

- No unbounded list or polling behavior.
- Authorization filters remain in optimized queries.
- Performance claims include workload, environment, and measurement method.

## Completion criteria

The bottleneck is measured, the change improves the target without regression, and results/limitations are recorded.