Agentic workflows and AI operating layers
AI command centers, automation workflows, intelligent agents, AI-assisted engineering, and approval-gated systems built around real operational work.
AI, ITAM, IT Security & Database Systems
I build practical AI systems on top of 22+ years of enterprise data and ITAM experience.
I combine senior enterprise software, database, ITAM/APM, SBOM, and secure automation experience with current hands-on work in agentic AI systems, AI operating models, and private business applications.
Independent AI systems practice
Dariusz Czechowicz
AI systems, ITAM intelligence, and enterprise data platforms
AI
AI
Agents
Data
ITAM
ITSEC
DB
22+
years in enterprise software, data, and platforms
3
specialist lanes: AI systems, ITAM intelligence, databases
Human
approval-gated AI systems, internal tools, and automation workflows
Expertise
A senior enterprise foundation, now focused on practical AI systems.
Software intelligence, SBOM, and risk data
Deep background in IT asset management, APM-adjacent data, software recognition, SBOM automation, vulnerability mapping, governance, and secure platform automation.
Database and data platform architecture
SQL Server expert with production experience across PostgreSQL, Oracle, MySQL, ETL, APIs, performance tuning, reliability, high availability, and data engineering.
AI Systems
Private apps and AI operating systems, protected by authorization.
The internal application list is not shown on the public page. Authorized users can access a protected systems view through the private gateway.
Advisory & Systems Work
Contact me when AI needs to become working infrastructure.
AI Workflow Redesign
Map an existing workflow, identify where AI can remove friction, and design a practical prototype with clear review gates.
ITAM / ITSEC Data Intelligence
Connect software, asset, SBOM, vulnerability, lifecycle, and operational data into systems that support decisions.
Database & Platform Modernization
Architecture, SQL performance, reliability, automation, APIs, data pipelines, and modernization for data-heavy platforms.
Selected work
Case studies — what I actually built, and what it produced.
Public, anonymised summaries of systems I designed, built, and operate. Each one maps a real problem to a working AI or data system. Full write-ups include architecture, decisions, and metrics.
Voice AI · 2026
Voice cloning TTS benchmarks on consumer hardware
Compared PocketTTS 2.1.0, OmniVoice-Triton, and VoxCPM2 (official + streaming) on a MacBook M4 and a WSL2/RTX rig. Built a reproducible benchmark harness for latency, throughput, and quality so engineering teams can pick a voice stack with evidence instead of vibes.
- 3 engines × 2 hardware targets
- ~10× latency gap, real-time vs offline modes
- Local-only — zero voice data sent to cloud
AIOS — a personal AI command center for real work
Built and operate an internal AI operating system that coordinates agents, tasks, approvals, triage, and revenue-facing workflows behind a single approval-gated surface. Same architecture pattern I now apply to client engagements.
- Approval-gated every external action
- Multi-agent routing + handoff
- Local knowledge + voice-oriented work surfaces
Q4C Estimation Assistant — BOQ extraction from tender PDFs
A document intelligence workflow that reads construction tender packages, extracts the bill of quantities, matches it against basis documents, and surfaces risk notes for the estimator. Built because BOQ estimation is slow, expensive, and error-prone in B2B construction.
- PDF → Excel with risk annotations
- Basis matching against reference docs
- Prototype lane, demoable for pilots
ReAImage — a private image-generation platform
Designed a private image-generation platform that runs local diffusion models behind an approval flow. Built for teams that need image AI without sending customer or product data to third-party generators.
- Local models, no external calls
- Approval flow on every output
- Pluggable model backends
h0st — lightweight AI service deployment
A lightweight deployment toolkit for spinning up isolated AI services (TTS, image, agents) on constrained hardware. Designed so that a single Mac mini can run several private AI surfaces without container sprawl.
- One machine, many isolated services
- Reverse-proxy friendly (Cloudflare Tunnel, nginx)
- Reproducible service manifests
About
After 22+ years in enterprise systems, my next chapter is building AI systems and applications.
After more than two decades building enterprise software, databases, ITAM platforms, automation, and data-intensive systems, I am focused on self-directed AI work: building my own AI operating systems, applications, and consulting patterns for companies that need AI to produce measurable value.
Contact
Contact me if you need AI redesign, ITAM intelligence, or database systems expertise.
FAQ
Frequently asked questions
What kinds of engagements do you take on?
AI workflow redesign (2–3 weeks), agentic prototype sprints (4–6 weeks), and embedded AI lead engagements (3+ months). Engagements usually start with a discovery call and a written scope.
Do you work on-site or remotely?
Mostly remote-first across EU time zones. On-site workshops are possible for engagements in Poland or with travel covered by the client.
Which languages do you work in?
English and Polish. The site itself ships in English; a Polish version is on the roadmap.
Can you sign an NDA before we share materials?
Yes — a mutual NDA is a standard first step before any discovery work that touches customer data, source code, or proprietary benchmarks.
What does "approval-gated AI" mean?
AI systems I design always require explicit human approval before they act on the outside world (sending messages, writing to production, changing infrastructure). The agent proposes; the human disposes.
Can I see your code or live systems?
Public case studies and anonymised architecture summaries are linked above. Source code and live systems are available under NDA or for active clients.