Steve Leggett

Performance engineering for responsive, reliable systems.

I help engineering teams find hidden performance constraints, understand their effect on users, and validate changes with evidence.

Explore case studies · Discuss a performance problem

How I work

I investigate latency, throughput, and resource limits across the UI, API, and database. Profiling and workload tests turn symptoms into a diagnosis—and a change we can measure.

Start here: Measure the Workflow, Not Just the Request — a blank contract for defining what your performance result includes.

From symptom to evidence:

• Trace the workflow. Find where users wait and where execution time goes.
• Name the constraint. Use profiling and workload evidence to understand what limits the system.
• Validate the change. State the expected outcome, measure the result, and check for regressions.
• Make it repeatable. Turn the diagnosis into useful tooling and checks the team can maintain.

Writing

Practical guides for investigating performance. Browse all writing.

• Before a Performance Fix, Write the Prediction

A reusable decision record for expected improvements, possible regressions, and the evidence that would challenge a fix.

• The Teardown Stack

A repeatable path from a slow-system symptom to a testable explanation.

• Data Growth Is Load: What Request Rate Leaves Out

A controlled comparison worksheet for investigating data growth at an unchanged request rate.

Case Studies

• 450ms → 45ms: finding a certificate-validation bottleneck

Certificate-validation latency: approximately 450 ms → 45 ms, as reported. Profiling exposed a caching gap in an enterprise desktop workflow.

• When a Green SLO Missed a Workflow Stall

Reported worst-case latency: approximately 2 s → 0.2 s. A conditional async path blocked a thread pool outside the existing measurements.

Browse all case studies

Work with me

Bring me the workflow that feels slow, the load test that leaves questions unanswered, or the performance change your team needs to validate.

Describe the workflow. What are users trying to do, and where do they wait?

Name the conditions. When does the problem appear, and what has changed?

Share the evidence. What have you measured or tried so far?

Describe your performance problem

LinkedIn · GitHub