Cloud & Infrastructure
Pragmatic architecture, migrations, reliability, and cost-aware operations built around the business—not the tooling.
Independent technology advisor · Poznań, Poland
Security, privacy
and AI governance.
I help leaders turn cloud, security, DevOps, and emerging AI into systems that are resilient, responsible, and ready for the real world.
Research terminology: on this site, an AI agent harness is the vendor-neutral control and execution environment around models and agents: context, tools, permissions, state, decision gates, evidence and effects. It does not refer to Harness.io.
Understand AI agents. Then go deeper.
Start with the complete L0–L8 architecture, then choose a practical security or GRC path.
AI Agent Harness Architecture: L0–L8 Map and End-to-End Example
Connect purpose, portfolio, runtime, context, tools, security, evaluation and execution through one access-review example.
Open the Knowledge Hub ↗AI Agent Security: Permissions, Prompt Injection and Safe Tool Use
How to secure AI agents across tools, identity, context and memory. A practical control checklist with failure tests and links to detailed architecture.
Read the overview ↗Agentic GRC: How to Use AI for Controls, Evidence and Assurance
A practical introduction to AI-assisted governance, risk and compliance: evidence collection, control mapping, human decisions and a bounded pilot.
Read the overview ↗Latest articles
Five source-attributed findings from verified Harness Collection PRs, with implementation limits.
Read the latest article ↗L5 · Applied controls · 5 Nov 2026Skills, MCP and Tools: From Discovery to Governed InvocationQualify capabilities, enforce per-action authority and verify effects.
Read article ↗L5 · Architecture · 3 Nov 2026Capabilities Are Not PermissionsWhy discovery and connectivity never grant authority to act.
Read article ↗Harness Weekly · 24 Oct–1 Nov 2026Harness Weekly: Evidence Before EffectsMonitoring, identity, skills, security evidence and privacy controls from the 1 November update.
Read article ↗Cloud Infrastructure ✦ DevSecOps ✦ Information Security ✦ AI Governance ✦ Resilience
Harness Catalog · Independent advisoryChoose the agent stack.
Choose the agent stack.
Keep the decision yours.
Vendor-neutral evaluation of AI agent platforms, controls and deployment evidence. Compare the options in your environment, challenge their boundaries and document an accountable go/no-go decision.
01 / Select
Fit before features
Compare interoperability, deployment constraints, security, privacy and total cost.
02 / Challenge
Test the boundaries
Define authorized negative tests for tools, identity, permissions and fail-safe behavior.
03 / Decide
Evidence before approval
Record observed results, gaps, ownership, residual risk and a defensible next step.
From strategic decisions to the operating details that make them work.
↗
DevOps & DevSecOps
Delivery systems that make secure defaults, clear ownership, and operational feedback part of everyday engineering.
Information Security
Risk-led security programs, DPO perspective, resilience planning, and controls that teams can actually sustain.
Responsible AI Agents
Least privilege, human approval, auditability, and practical guardrails for AI systems that can take action.
How I work
Clarity before complexity.
The strongest technology decisions happen when business context, operational reality, and risk are considered together. That’s where I work best: across the boundaries.
01
Business first
Translate technical choices into cost, risk, speed, and resilience.
02
Secure by design
Build ownership and controls in from the start—not after launch.
03
Useful over novel
Choose technology for the outcome, not for the trend cycle.
20+years across IT operations, cloud, and security
4advisory domains: cloud, security, AI, and GRC
Publicresearch notes with claim boundaries, source links, and explicit evidence limits
Trusted practiceSecurity, ethics
Security, ethics
and responsible innovation.
Independent advisory grounded in practical protection, cloud security, and accountable AI.
CSA AI Trustworthy Pledge. Voluntary public pledge. Participation is listed by CSA under DPO Aleksey Lekontsev. This is not a certification.
Cybersecurity Awareness Month 2026 Champion. Voluntary participation in the National Cybersecurity Alliance Champion program; the program is free to join. This is not a certification.