12+ YEARS OF OPERATOR EXPERIENCE

E-COMMERCE
INFRASTRUCTURE
& AI DELIVERY

SoftSpark is built and led by Łukasz Krzemień.
We design, build, and stabilize revenue-critical platforms
across commerce, infrastructure, production delivery, and AI systems.

FOUNDER & ENGINEER

ŁUKASZ
KRZEMIEŃ

Łukasz is a software and infrastructure engineer who works directly on architecture, implementation, delivery, and production operations.

He combines 12+ years of commerce-platform experience with infrastructure automation and AI engineering. Important systems need explicit ownership, measurable acceptance criteria, repeatable delivery, and a tested recovery path.

ROLE Architecture through operations Founder-led means the person defining the system also works on its implementation, release path, observability, and recovery.
COMMERCE Magento / PHP / PWA Catalog, checkout, search, cache, PIM, integrations, and multi-store operations.
INFRASTRUCTURE Cloud / IaC / CI/CD Provisioning, containers, delivery pipelines, observability, backup, and recovery.
AI SYSTEMS RAG / MCP / Agents / Orchestration Knowledge retrieval, coding-agent platforms, model routing, permissions, evaluation, and human approval.
INFRASTRUCTURE

THE SYSTEM
BEHIND THE
REVENUE

We operate in the layer where uptime, release flow, observability, and platform decisions directly shape commercial outcomes.

Magento / PWA / Akeneo Symfony / Pimcore AWS / Azure / Hetzner CI/CD / IaC / Containers
CORE CONTROL 99.95% stability target across release-critical paths
Provisioning Terraform / Terragrunt Hetzner Cloud, AWS, and hybrid VPS setups
Automation Containerized Ansible Repeatable provisioning and server management
Delivery CI/CD with zero-downtime Validation, rollback readiness, and zero-downtime release patterns
Security Access, secrets, backups
Recovery Rollback ready
Observability Full-stack visibility

OUR
APPROACH

Operator mindset

We care about uptime, release safety, recovery, and delivery speed, not presentation theatre.

Automation first

We replace tribal knowledge with Terraform, Ansible, Docker, CI/CD, runbooks, and repeatable workflows.

Commerce-native thinking

We understand catalog, checkout, search, queues, cache layers, integrations, and peak traffic realities.

AI with guardrails

We use AI where it shortens lead time and makes knowledge reusable, but keep humans in control of decisions.

WHAT WE HELP TEAMS FIX

Releases that are too risky because the platform depends on heroics.

Infrastructure that grew fast but was never standardized.

Magento or PWA performance problems that leak revenue and slow product work.

Missing operational discipline: poor observability, weak rollback paths, undocumented flows.

AI experimentation without governance, traceability, or integration into real delivery work.

E-COMMERCE

REVENUE
ENGINE

We work directly in the application layer where performance and stability affect revenue, from catalog and checkout to search, cache, feeds, and multi-store operations.

Checkout layer

Speed where it converts

Magento 2 platform engineering for B2C and B2B commerce. Headless and PWA storefront architectures with intelligent middleware proxies.

APM TTFB CVR
Search / Cache

OpenSearch + Varnish aligned

Varnish, OpenSearch, queue, and cache tuning for high-traffic stores.

PIM / Data flow

Akeneo PIM integration

Akeneo PIM integration and data flow design.

Ops coverage

Support that sees the whole system

  • Multi-store architecture
  • Production diagnostics
  • Traffic-ready releases
  • Monitoring + incident loop
AI DELIVERY

AI DELIVERY
SYSTEMS

We treat AI as an engineering system, not a chat window. Rules, permissions, retrieval, evaluation, and release gates remain explicit.

CONTROL PLANE Rules + Retrieval + Delegation least-privilege tools, approval gates, evaluation, and human review
ENGINEERING TOOLKIT ai-toolkit 109 skills, 44 agents, hooks, audits, and shared policies across 12 developer-tool integrations
KNOWLEDGE RAG + CRAG + multi-hop + vision RAG-based knowledge bases with semantic search, CRAG, multi-hop reasoning, and vision support
WORK MANAGEMENT jira-mcp Multi-instance Jira access with ADF conversion, local caching, and reusable templates
HOW WE WORK

ENGINEERING
PROTOCOL

01

Diagnose

Audit architecture, release flow, performance risks, operational pain points, and where the team is losing speed or control.

02

Design

Define the architecture, automation boundaries, environments, observability, and measurable success criteria.

03

Codify

Codify infrastructure, containerize services, standardize pipelines, and document the path so the team can repeat it.

04

Ship

Small reversible changes, validation-first workflows, rollback readiness, and clear operational checkpoints.

05

Transfer

Monitor, tune, document, and leave behind SOPs, runbooks, and tooling that survive beyond one release window.

WORKING
STYLE

Direct and measurable

Clear goals, clear trade-offs, clear ownership.

Strategic but hands-on

Architecture decisions backed by implementation.

Three options, not one guess

We compare paths before committing.

Documentation as leverage

If a process matters, it should be reusable.

Focus on the real bottleneck

Not the loudest symptom.

TYPICAL OUTPUTS

Architecture

Target topology, migration plan, dependency map, risk list

Implementation

Terraform modules, Ansible automation, Docker images, CI/CD changes

Validation

Syntax checks, deployment verification, rollback path, health checks

Operations

Monitoring, alerting, runbooks, cost and performance improvements

Enablement

SOPs, KB articles, AI assistant workflows, team-specific tooling

OPEN SOURCE

PUBLIC
REPOSITORIES

Tools we build and use in delivery, published under Apache-2.0.

ai-toolkit
Shared AI engineering toolkit with 109 skills, 44 agents, hooks, policies, audits, and 12 developer-tool integrations.
jira-mcp
Multi-instance Jira MCP server with ADF conversion, local caching, bulk operations, and reusable templates.
gitspace
Oh My Zsh plugin that binds Git identity to workspace directories, including commit email, GitHub account, SSH key, optional SSH signing, and identity auditing.
mage2x
Runs Magento and shell commands inside Docker, Podman, or Kubernetes workloads. Refuses ambiguous targets and confirms destructive production operations.