Zen Solutions builds AI systems that can ship.

The work starts with a practical question: what can a team actually run, inspect, improve, and own? Everything else is downstream of that.


Zen Solutions artwork: a crimson bonsai before a glowing moon over a dark lake.

Hans Havlik

AI Solutions Architect

Architect. Builder. Student.

Developer and AI/ML systems architect with 10 years as a technical builder, 4 of them in enterprise AI delivery. I have owned agentic orchestration work through full development cycles: design, architecture, build, and deployment. Those systems ran in aerospace, global logistics, healthcare, financial services, IT, knowledge management, and project management.

I trained as an EMT at sixteen. From 2017 I wrote backend services and operational automation for a hosting company. I worked the pandemic years at the Malcom Randall VA Medical Center. In 2022 I moved into enterprise AI delivery. The career was built from scratch, in that order, and none of it was planned.

I am a practising Gaudiya Vaishnava, in a devotee community in Alachua, Florida. The practice is older than the career, and it shapes what I will and will not build.

The most important thing in this moment is not the AI systems. It is the people who use and direct them: their understanding, their judgment, and the human connection that no system can replace. Zen Solutions is built on that principle.

Build with intention. Scale with clarity. Keep the keys in the client's hands.

Architecture is only useful when the system keeps working after handoff.

The question underneath every engagement is who holds the judgment. Which calls the model is allowed to make, who checks its output before it reaches anyone, and who is accountable when it is wrong. Those answers are the architecture. The rest is implementation.

Most AI work gives that away without meaning to. A platform gets bought before anyone writes down the problem it was meant to solve. The team ships what the model said. Nobody is left who can argue with it. Every engagement here leaves the deciding with your people.

01

Agentic AI systems

Production-minded agents designed around real tools, state, memory, recovery, and human oversight.

02

Cloud AI and DevOps

Deployment paths, infrastructure as code, observability, and operational ownership for systems that need to run reliably.

03

Automation and integrations

Data pipelines and workflows that connect fragmented tools, trigger useful actions, and reduce manual operational drag.

04

Web, hosting, and business systems

Websites and business infrastructure that clients own, control, and can maintain without being trapped by a vendor.


What I work in.

The tooling landscape evolves constantly. This is the stack as of October 2026, built through hands-on delivery across automations, application integrations, and deployed agentic workflows, in production.

AI & agentic engineering

  • Azure AI Foundry
  • Azure OpenAI
  • Microsoft Agent Framework
  • LangChain
  • LangGraph
  • MCP
  • FastMCP
  • RAG
  • Guardrails
  • Context engineering

AI-assisted development

  • Claude Code
  • OpenCode
  • Cursor
  • GitHub Copilot
  • Spec-driven development
  • Git worktrees
  • Reviewer-agent workflows

Backend & full stack

  • Python
  • FastAPI
  • Flask
  • Pydantic
  • SQLAlchemy
  • TypeScript
  • Node.js
  • Next.js
  • React
  • Astro
  • Vite

Data & retrieval

  • Azure AI Search
  • PostgreSQL
  • Supabase
  • pgvector
  • Redis
  • Firestore
  • SQLite FTS5
  • Hybrid & vector retrieval
  • Row-level security

Cloud, DevOps & security

  • Azure
  • Google Cloud
  • Cloudflare
  • Fly.io
  • Docker
  • Bicep
  • Terraform
  • GitHub Actions
  • Azure DevOps
  • Secret scanning
  • Least privilege

Observability & ops

  • OpenTelemetry
  • Azure Monitor
  • Application Insights
  • AppDynamics
  • Grafana
  • Deployment validation

Enterprise platforms

  • ServiceNow
  • Microsoft Teams bots
  • ITSM
  • Jira
  • SAP
  • Obsidian

Automation & integration

  • n8n
  • Microsoft Graph API
  • Meta Graph API
  • Apify
  • Firecrawl
  • REST APIs

The skill crosses over. The business does not.

Zen Solutions is an independent practice and its scope is deliberately narrow: small businesses, individual builders, and my own products. It is not a competing consultancy, it does not pitch for enterprise consulting work, and it does not solicit clients from anywhere. The enterprise delivery behind my experience stays unnamed on this site on purpose, described by sector and architecture. That is a confidentiality obligation I intend to keep.

What moves between enterprise delivery and this practice is skill, not business. It runs in a circle, and the circle is the actual point.

  1. 01

    Build it here first

    On my own time and my own risk, where a bad decision costs me and nobody else. A pattern I have not run myself is a pattern I have no business recommending to you.

  2. 02

    Prove it where it counts

    Under real constraints, on data that is not mine, with review I cannot waive. That is where a pattern either holds or quietly comes apart, and you only find out which at that scale.

  3. 03

    Write down what survived

    Including what did not, which is the more useful half. Zen Agent Skills is exactly this: the patterns that held, in a form something else can run.

  4. 04

    Hand it on

    To colleagues, to whoever reads this site, to friends who are learning, and to my son on a Minecraft server where the stakes are low and the debugging is entirely real.

Every method in the library has been run on real work, broken, and run again. The library is free. The practice is for hire. Both rest on that loop.


An applied R&D practice, in the open.

Alongside the enterprise work I keep an active independent engineering practice: agentic systems, AI-assisted delivery, backend and full-stack builds, cloud deployment, security controls, and hardened CI/CD. It is where emerging orchestration, retrieval, verification and human-in-the-loop patterns get tested before I would put them anywhere with consequences.

Most of that project code was produced by AI coding agents working under a written specification, with architecture direction, review, approval, integration and operational oversight from me. I say that plainly because it is the method I sell, it is what Zen Agent Skills exists to make repeatable, and a practice that works this way should be willing to admit it.

  • Checkpointed LangGraph workflows with guardrail retries, human-review interrupts, and resumable execution.
  • Governed tool access through FastMCP services, exposed to coding agents and AI clients under typed contracts.
  • Hybrid retrieval over lexical and vector search with rank fusion, confidence routing, and cost circuit breakers.
  • Evidence that the tests work, using revert probes to show a new test actually detects the failure it claims to.
  • Decisions kept in architecture decision records, task ledgers, and after-action reviews, where the next person can find them.

The public work teaches people to build.

Zen Solutions publishes practical tutorials, architecture breakdowns, long-form writing, and open-source tools. The free AI Practitioner's Field Manual is the starting artifact for builders who want a grounded path through the current AI landscape.

Ready to discuss the work?

Start with the problem, the operating context, and what must still work when the prototype becomes a real system.