Consulting

JVM and low-latency systems performance consulting.

I work with engineering teams on high p99 latency, throughput ceilings, memory pressure, GC pauses, lock contention and rising infrastructure cost — through targeted investigations, architecture reviews and hands-on training. Every engagement starts with a conversation about the actual problem, not a fixed package.

Ways to work together

Four engagement types

Each describes when it fits, what it covers, and what you get at the end. None of these are fixed-price packages — scope is agreed before work starts.

INVESTIGATION

JVM performance audit

When it fits: you have a specific performance concern — high p99, a throughput ceiling, unexplained GC pauses, rising CPU — but no clear diagnosis yet.

What it covers: workload analysis, profiling, allocation and GC review, contention analysis and JIT/compilation behaviour, against a representative workload.

What you get: a written diagnosis of the actual bottleneck(s), evidence behind each finding, and a prioritised list of remediation options — not a guaranteed percentage improvement.

ARCHITECTURE

Low-latency design review

When it fits: you're designing or reworking a latency-sensitive component and want a second, specialised opinion before or during implementation.

What it covers: execution model, state ownership, communication paths, persistence boundaries and failure semantics, reviewed against your actual latency and throughput targets.

What you get: a written review — risks, trade-offs and concrete recommendations tied to your constraints.

PRODUCTION

Performance intervention

When it fits: something in production has regressed or hit a wall — a latency regression, throughput collapse, memory pressure, high CPU, lock contention or a scaling limit — and you need focused help resolving it.

What it covers: targeted diagnosis on the affected path, working from your existing telemetry and profiling where available, adding instrumentation where it's missing.

What you get: root-cause findings and a concrete fix or remediation plan your team can implement and validate.

ENABLEMENT

Advanced engineering training

When it fits: your team is capable but hasn't built deep intuition for JVM internals, concurrency or performance-conscious design, and would benefit from structured, hands-on sessions.

What it covers: JVM internals, concurrency and memory ordering, profiling and benchmarking methodology, and performance-conscious architecture — tailored to your codebase and team level.

What you get: working material and exercises your team keeps, built around problems close to your own systems.

Process

How an engagement typically runs

01

Discuss the problem

A conversation about what's happening, what's been tried, and whether this is a good fit.

02

Agree scope and access

What's in scope, what data or environment access is needed, and what a useful outcome looks like.

03

Analysis

Profiling and investigation against the agreed scope and a representative workload.

04

Recommendations and validation

Findings, prioritised recommendations, and an agreed way to validate the result.

Engagement principles

Scope and success criteria are agreed before work begins; recommendations are based on measurements from your workload.

Get in touch

Start a conversation

Email contact@kzybala.pl with a short description of the problem.

Useful to include, if you have it:

  • the symptom — what's slow, unstable or expensive, and how you noticed
  • tech stack and runtime (JVM version, language, key frameworks)
  • rough scale — request rate, data volume, latency/throughput targets
  • what a good outcome would look like

Please don't send production data, credentials or anything confidential in a first email — what to share, and how, gets agreed once we're talking.