// AI Architecture & Consulting

Your pilot works.
Production is the other 90%.

I help engineering and product teams take AI agents the rest of the way: the architecture, the deployment path, the evaluation loop that proves it works, and the operating discipline that keeps it working after I hand it over.

Book a 30-minute scoping call See how engagements work

Start with a $2,500 one-week audit  ·  Read the Field Manual first

Scroll

// Why this is hard

The demo was never the hard part.

Getting an agent to work once is a weekend. Getting it to work reliably, under audit, on data you do not control, with a way to prove it still works after the model changes underneath you, is the actual job. Most teams are not stuck because they picked the wrong framework. They are stuck because nothing downstream of the demo was ever built.

If any of that sounds like your situation, the work below is what I do about it. If your agent has not been built yet, start at the audit. If it has been built and nobody can tell you whether it is working, start at the harness.

// What I do

Start small. Decide from evidence.

One week, one fixed price, and you own everything it produces whether or not we work together again. If you do continue, the fee credits against the next phase. The larger work is scoped against what the audit actually finds, because quoting a build before anyone has looked at your data is how both sides end up wrong.

01 —

Agent Readiness Audit

1 week$2,500 fixedfee credits forward

For teams who have decided to build something with AI and want to know what they are walking into before they commit budget to it.

  • System and data map. What exists, where the data lives, what an agent can actually reach, and what is blocked by access, quality, or policy.
  • Scored use-case register. Every candidate ranked on value, feasibility, data readiness, and risk, including the ones you should not build and why.
  • Acceptance criteria for the top use case. Input classes, expected output properties, a failure budget, and the check that proves it. This is the part almost nobody hands you, and it is what makes the next phase reviewable.
  • A 30-day plan with an owner and exit criteria per step, plus a recorded 60-minute walkthrough you can forward to whoever holds the budget.

Outcome: you can brief your own team, defend a plan to the people funding it, and start building against criteria instead of vibes. If the honest answer is that you should not build it yet, you get that too.

Written artifactsNot a slide deckYours to keep

02 —

Agent Pilot Sprint

4 to 6 weeksscoped from the audit

For teams with an approved use case who need it built properly the first time.

  • A working agent against real tools and real data, not a scripted demo.
  • State, memory, and recovery designed for the failure modes you will actually hit.
  • An evaluation set and a pass criterion agreed before the build starts, so "is it working?" has an answer that is not a hunch.
  • Tracing wired in from day one on OpenTelemetry GenAI conventions, so you are not locked to one observability vendor.
  • A deployment path with repeatable releases, a documented rollback, and a runbook written for whoever operates it after me.

Outcome: something in front of real users, instrumented well enough that the next decision is informed by data rather than opinion.

Multi-Agent DesignTool Integration Evals & AcceptanceObservability

Built with LangGraph · LangChain · Microsoft Agent Framework · Google ADK · MCP · Claude Code · OpenCode

03 —

Embedded AI Architect

monthlyscoped to your teamcancel monthly

For teams with capable engineers who want someone who has already made these mistakes sitting alongside them.

  • A fixed number of days each month, named up front, on your architecture and your review queue.
  • Design review, code review, and the unglamorous decisions: model routing, cost ceilings, context strategy, what to cut.
  • Your engineers learn the method rather than renting it. That is the point, and it is why these engagements are meant to end.
  • No minimum term, no exit fee, no leverage held over you.

Outcome: your team ships faster and needs me less each month, which is the only honest definition of a retainer working.

Architecture ReviewDesign Partnership Team CapabilityCancel Monthly

Founding client terms, stated plainly. Zen Solutions is new as an independent practice, and the enterprise work behind it sits under NDA. So the first engagements are priced below what this work is worth, in exchange for one thing: permission to publish a written case study with real figures once it ships. You get senior delivery at an early-stage rate. I get proof I am allowed to show you. When the case studies are on this page, the pricing goes up.

// Applied across all work

Governance & Compliance Documentation & SOPs Knowledge Management Agent Observability Performance Monitoring Process Design

You keep the keys.

Small business tech has a dependency problem. A vendor builds your site on their platform, registers your domain in their account, and holds every login. When the relationship sours or they get complacent, you are a tenant in your own digital property, and you have no leverage.

That is not how I work. The domain lives in an account you control. The hosting is yours. The login to edit your own site belongs to you. If the relationship ends, nothing critical goes with me, because nothing critical was mine to begin with.

The job is to build something that runs well, belongs to you, and keeps running after I am done. Anything past that is a choice you make, not a necessity I manufacture.

This site is built and hosted by Hans on Cloudflare. If you want to know how, that is a conversation we can have.

Book a 30-minute scoping call

// Selected delivery

Three systems, in production.

Enterprise delivery under NDA, so the clients are described by sector rather than named. What is described is the architecture, the constraint that shaped it, and the part I personally owned.

Global logistics & enterprise IT · AIOps

Agentic operations platform for SAP and ServiceNow

Situation

Alert volume across SAP and ServiceNow was outpacing the operations team. Triage was manual, inconsistent between shifts, and the institutional knowledge for handling an incident lived in people rather than in the system.

Constraint

Global operations, so nothing could be a black box. Every automated action needed to be traceable after the fact, and a human had to stay in the loop on anything consequential.

What I built

A Microsoft Agent Framework platform handling SAP alert processing, ServiceNow workflow automation, AI-assisted incident triage, and RAG-backed recommendation workflows, instrumented end to end with OpenTelemetry.

What I owned

Agentic orchestration through the full development cycle: design, architecture, build, and deployment.

Result

Deployed and running across global operations.

Enterprise IT · Infrastructure automation

Natural language to reviewed, deployable infrastructure

Situation

Infrastructure requests queued behind a small number of engineers who could write Bicep and Terraform correctly. The bottleneck was not intent, it was translation and review.

Constraint

Generated infrastructure code cannot be trusted on sight. Nothing could reach an environment without security review and an explicit human approval.

What I built

A multi-stage pipeline: natural-language understanding, code generation, automated security review, a GitHub pull-request workflow, and a human approval gate, with WebSocket status streaming so requesters could watch their request move rather than wait blind.

What I owned

Platform architecture and the agentic pipeline, including the review and approval gating.

Result

Requests arrive as reviewable pull requests rather than tickets.

Enterprise IT · GenAI service management

The assistant that became the platform

Situation

Incident management, SLA tracking, and notification handling were spread across tools and people, with no single surface an engineer could ask a question of.

Constraint

It had to prove its value on real ITSM workflows before anyone would fund a platform build on top of it.

What I built

A LangChain work assistant covering incident management, ServiceNow workflows, SLA tracking, and autonomous notification handling.

What I owned

The GenAI foundation, end to end. It seeded the Azure-native platform above, which now serves global enterprise operations.

Result

Earned the mandate for the larger platform.

Numbers matter more than narrative, and I would rather leave a gap than fill it with something I cannot source. The slots above are marked because the figures are still in NDA review, not because they do not exist.

Book a 30-minute scoping call

// Open Source

How I work, in public.

AI coding tools now produce more code in an hour than a person can read in a day. The usual answer is to read every line, which makes the reviewer the bottleneck and quietly caps how much a team can take on.

Zen Agent Skills is the method I use on client work, released free and open source so you can read it before you hire me. Twenty written playbooks that give an AI assistant a consistent way to work: agree the contract in writing before anything is built, then verify the result with evidence after. Plain Markdown and a few standard-library Python scripts. No platform, no service, no account.

This is the same discipline behind the audit and the harness above. The written specification you approve, the acceptance criteria you can check, the verdict that admits when it cannot be sure: those are not marketing language, they are the artifacts, and the repository is the proof that they are real.

MIT Licensed20 Skills Python 3.11+Claude Code CursorVS Code / Copilot OpenCodeNo Dependencies

// How it works

01 —

Agree the contract

Before anything is built, the idea becomes a short written specification: what it should do, what it deliberately should not, and how you will know it worked. You approve it. Nothing gets coded against a draft.

In practice: the expensive disagreements happen on one page, not inside a finished feature.

02 —

Break it into small jobs

The approved specification is decomposed into atomic tasks, each small enough to finish, check, and undo on its own, each carrying the command that proves it is done.

In practice: progress arrives in units you can actually review.

03 —

Build in isolation

Independent tasks go to separate agents, each working in its own copy of the project so they cannot trip over each other. Results are reconciled deliberately, not merged on faith.

In practice: parallel speed without the parallel mess.

04 —

Verify with evidence

A final independent pass runs the declared checks, audits the result against the specification, and returns pass, fail, or blocked, each with the evidence behind it.

In practice: "it works" becomes a claim someone can check, not a status update.

What's in the box

  • Set up a project and its work tracker so an agent can work in it safely
  • Turn a rough idea into a written specification you approve first
  • Gate the plan, and refuse to start when it is not yet implementable
  • Break approved work into small, self-contained, verifiable jobs
  • Run several jobs in parallel, isolated, then reconcile them on purpose
  • Derive the tests from the specification, not from the code that was just written
  • Write the docs, the pull request, and the handoff so the next person is not guessing
git clone https://github.com/hams-ollo/zen-agent-skills.git
cd zen-agent-skills
python scripts/install.py --dry-run

The dry run prints exactly what would be installed and where, and changes nothing. Drop the flag when you are ready.

It tells you when it cannot be sure.

Most AI workflows report success, because reporting success is easy. This one returns one of three verdicts, and the third is the point. blocked means the check could not run honestly. Here is a real one from the repository, dated 2026-07-27, not an illustration:

verdict: blocked
blocking_reasons:
  - reason: the supplied contract is not approved
    detail: docs/spec/house-review.md carries `status: draft`.
      A draft spec is one no human has agreed to, so verifying
      against it would launder an unapproved contract into evidence.

That specification has since been approved. The record is left dated rather than quietly updated, because a verification that was true last month is not evidence about today.

↗ View on GitHub Read the Founder's Guide

// Knowledge & Community

I also teach this, for free.

Free access to the knowledge that matters. The goal is never gatekeeping: it's building a community of people who actually understand the tools they're working with.

Give someone a solution and you help them once. Teach them to build solutions and you change what they are capable of.

Deeper content, structured courses, and a community for those who want to go further, for a fair price.

YouTube

Zen Solutions

Tutorials, breakdowns, and real builds. AI architecture explained without the fluff.

Free

Substack

Zen Solutions

Long-form writing on AI systems, architecture decisions, and lessons from the field.

Free

Patreon

Zen Solutions Community

Courses, webinars, and a community of builders. Structured learning for those going deep.

Members

GitHub

Zen Solutions

Open source tools, agent frameworks, and production code from the field.

Free

// Free Resource

The AI Practitioner's Field Manual

The reference I wish someone had handed me three years ago, now in its third edition. A grounded, honest walkthrough of the AI landscape in 2026: what is actually possible, what is overhyped, what is dangerous to skip, and a 90-day path to becoming a real practitioner.

Not a prompt dump. No ten ChatGPT hacks. Written from years of production work at the intersection of enterprise AI and the discipline of building well.

What's inside

  • Where we actually are: capability, context, and agency
  • The four practitioner levels: User → Builder → Operator → Strategist
  • Two techniques that compound: Analogical Translation & Reverse Prompting
  • The five disciplines: context, restraint, verification, security, leadership
  • When not to use AI, and why most AI projects fail
  • Setting up your workbench: GitHub, git, editor, terminal agent, permissions
  • Giving your agents real tools, and the portability problem
  • Security & governance — the part nobody covers
  • A concrete 90-day practitioner's path
Read the full manual — free, no form Download the PDF

What is new in version 3

Ten parts, from where the field actually is to the workbench you build on and the ninety days that get you there.

Version 3 adds the parts readers kept asking for: why most AI projects fail, how to stand up a real agent workbench from GitHub outward, how to give agents portable tools, and a governance baseline that is not a compliance lecture.

Version 3.0 · August 2026 · 30 pages · Free to share

// Selected Work & Experiments

Systems I've shipped.

Public platforms, private tools, collaborative builds, and field experiments. AI and systems work across learning, education, knowledge infrastructure, and community.

Collaboration · Education Technology

HT Sourdough

An agentic AI-powered homeschool education platform built around a single design principle: technology as a precision compass for analog living, not another screen to disappear into. Built for intentional families who want AI to serve their life, not consume it.

Hans contributes the knowledge infrastructure layer — Savitri's MCP server pattern adapted as a portable, student-owned knowledge garden. When a student leaves the program, they leave with a living knowledge base that is fully theirs: agent-accessible, portable, never locked to the platform.

↗ Platform

"The digital compass for analog living. Twenty minutes of focused screen time powers the whole day."

— HT Sourdough platform philosophy

A three-node collaboration: Lila Purushotama builds the learning platform, Madan (True North Learning Collective) grounds it in a real community with Florida state grant funding, and Hans builds the knowledge layer the student carries for life.

AI Platform · Co-founder

MindRev

An AI-powered coaching and learning platform built on LangGraph orchestration. Personalized coach agent, structured onboarding flows, and a knowledge architecture designed for compounding improvement over time.

↗ Live App ⌥ GitHub

Content Infrastructure · Open Source

Zen Solutions Studio

The AI development and content production workspace behind the Zen Solutions educational track. VS Code Nightly and Claude Code as the cockpit, Obsidian as the knowledge layer, n8n for automated capture and content pipelines. Zen Agent Skills is the open-source artifact the whole system produces.

↑ Zen Agent Skills ⌥ GitHub

Personal Tool · Knowledge Architecture

Savitri

A living second brain and personal knowledge management system built for daily use. Obsidian as the writing surface, Supabase and pgvector for semantic retrieval, a custom MCP server for agent access, and n8n automation for frictionless capture. The knowledge is portable, agent-accessible, and sovereign.

Private Preview

Community · Archery & Bhakti Yoga

Dhananjaya Dojo

A training program and session tracker for the local archery dojo in Alachua, FL, serving children and adults pursuing precision and discipline. Includes a session logging app with an integrated AI sensei grounded in Bhagavad-gītā As It Is — Arjuna's own text, as the source of guidance.

Local Program

AI Platform · Production Readiness

Ship It Right

An AI-powered production readiness and transformation advisory platform. Three assessment tracks — Builder, Business, and PM — run parallel 8-domain analysis grounded in a session Knowledge Base. Single-click Web Search Report via Gemini 2.5 Pro delivers cited, live-researched roadmaps with 90-day action plans.

Private Beta

Community Tool · Simulation & Tactics

Pucker Point

A ballistic calculator and fire mission management system built for the ARMA Reforger realism community. Handles call-for-fire workflows, range and deflection computation, and mission coordination for WCS server operations. Systems thinking applied to a domain well outside AI.

In Development

// Who I am

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

Hans

AI Solutions Architect

Architect. Developer. Student.

Developer and AI/ML Systems Architect with ten years as a technical builder, including five in enterprise AI delivery. I have owned agentic orchestration work through full development cycles, designing, architecting, developing, and deploying AI and ML systems across aerospace, global logistics, healthcare, financial services, IT, knowledge management, and project management.

My work sits at the intersection of enterprise AI architecture, years of clinical practice, and a Gaudiya Vaishnava framework I have lived from the inside rather than studied at a distance. I served as a Medic in the US Army, worked through the pandemic years in the Malcom Randall VA Medical Center, and built a technical career as a software developer and consultant from scratch. 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.

The tooling landscape evolves constantly. What follows is the stack I work in now, built through hands-on delivery across automations, application integrations, and deployed agentic workflows — in production.

AI & Agents

LangChainLangGraphMicrosoft Agent Framework Google ADKClaude CodeOpenCodeMCP

Cloud & DevOps

AzureAzure AI FoundryVertex AIFirebase DockerKubernetesTerraformLinux

Backend & Data

PythonNode.jsPostgreSQL RedisNeo4jSupabase

Automation & APIs

n8nMicrosoft Graph APIMeta Graph API ApifyFirecrawl

Observability & Ops

GrafanaAppDynamicsLightspeed

Enterprise Systems

ServiceNowJiraAtlassian SAPObsidian

Web & Hosting

CloudflareCloudflare PagesWordPress GhostDNSCMS Configuration

// Let's Build

Ready to ship
something that matters?

Whether you're building enterprise AI, exploring what's possible, or looking to learn: the conversation starts here.

Book a 30-minute scoping call Explore Free Tutorials

// Join the community

Discord Server