Michael Kaas · Germany
Personalized workflowsAI agentsWeb apps

Software your team willactually use

Web apps, AI workflows, and the integrations between them. Built one project at a time, by one specialist, for small teams that need software they can rely on.

4+
Years shipping
20+
Production systems
99.9%
Avg. uptime
Use cases

Still doing this manually?

Generic automations are everywhere. What you need is a hyper-personalized workflow built around your daily friction, not another template that almost fits.

For individuals

Still doing these manually?

  • Your morning spread across 4 apps, every single day
  • Receipts, files, and emails sorted by hand
  • Paid something twice and found out too late
  • Meeting done. Action items still in your head

I map the tools you already use, then build one workflow that handles it, with alerts when something breaks.

For store owners

Running a store but still doing this manually?

  • Logging orders, receipts, and refunds yourself
  • Chasing reviews after every delivery
  • Supplier out of stock and you find out too late
  • Managing multiple stores eating your whole day
  • Support tickets piling up faster than you answer

I connect your shop, inbox, and spreadsheets into one reliable flow, so ops run while you focus on growth.

Services

Three things I do well

Web apps and internal tools

Next.js apps, dashboards, admin panels, marketing sites with a real backend. The kind your team uses every day, not the kind that looks good in a screenshot.

  • Next.js front end, typed end to end
  • Backend, database, auth, the lot
  • Designed to be edited later, by future you or someone new
  • Deployed, monitored, documented

AI agents and workflows

Agents that read documents, draft replies, summarise meetings, or triage tickets. Built with an evaluation set, validation, and a human-review fallback so 'mostly works' never reaches your customers.

  • Agent built around your real data, not a generic template
  • Evaluation set so you know its accuracy before launch
  • Validation, retries, human review for low confidence
  • Plugged into Slack, Notion, your CRM, whatever you use

Hyper-personalized automations

Not another Zapier template. Workflows built around how you actually work. Make.com, n8n, and custom backend glue when no-code hits its limit.

  • Mapped from your real daily friction, not a generic checklist
  • Morning routines, inbox sorting, duplicate-payment alerts, post-meeting capture
  • Store ops: orders, reviews, stock alerts, multi-channel support triage
  • Error handling, alerts, and a runbook your team can edit
0+
Years building software
0+
Projects shipped
< 0h
Avg. reply to a new email
0%
Projects with a written runbook
Work

Recent work

AI & backend

AI systems and the backends they run on

Most of my work is building AI agents, model demos, and API integrations — then wiring them into Python and Node services, databases, and production workflows. Front-end is part of the delivery when a product needs a UI; the depth is in AI engineering and backend.

Every skill below is part of my active stack.Core= strongest professional focus.

AI & LLM engineering

Agents, demos, model tooling, and integrations that survive real data — with prompts, validation, and fallbacks baked in.

OpenAI APICorePythonCoreLangChainCorePrompt EngineeringCoreAI AgentsCoreRAG & Vector DBsAnthropic APIHugging FaceLLM EvaluationTool Calling

Backend & APIs

Services, webhooks, and REST layers that agents and apps depend on.

FastAPICoreNode.jsCoreExpress.jsREST APIsAPI IntegrationWebhooksStripe

Data & storage

PostgreSQLCoreMongoDBRedisSupabase

Automation & orchestration

n8nZapierMake.com

Front-end & UI

Next.jsReactTypeScriptReact NativeTailwind CSS

DevOps & delivery

DockerAWSGit
The process

How a project actually goes

  1. 01

    Initial contact

    You tell me what your team is trying to do and where the friction is. I ask questions to understand the work. Plain English, no pitch.

  2. 02

    Plan and quote

    I send back a written plan: what gets built, what's deliberately out of scope, the price, and a timeline scoped to your actual project. If it doesn't fit, we stop here. No obligation.

  3. 03

    Build in slices

    I build the work in vertical slices and share progress at regular check-ins. You can change direction at any point without it costing more.

  4. 04

    Hand-over and free fixes

    You get the runbook, a recorded walkthrough, and a follow-up period for free bug fixes after launch. After that, you can run it yourself or send the occasional task.

The difference

Why work with me instead of an agency

One specialist, not a team

You reach me directly. I do the work. I do the hand-over. No PMs, no junior devs, no game of telephone between you and the person actually building it.

Practical, not theoretical

I build things that ship. If something is risky or unlikely to work on your data, you'll hear about it before you've paid for the full build, not after.

AI that ships, not AI that demos

Every AI workflow I build comes with an evaluation set and a fallback path. If accuracy isn't good enough on your real data, you'll know before launch.

Written so your team can run it

Every project ends with a runbook plain enough that a non-developer can change a webhook URL, swap an API key, or read the logs when something looks off.

Michael Kaas
4+
Years Exp.
Profile

About me

Michael Kaas

I'm a software developer, which means I argue with machines that are always right until 4:47 p.m. on a Friday. I joined for meetup pizza, unlimited tabs, and saying "it works on my machine" while production quietly panics. I debug with a rubber duck, deploy with courage, and estimate tasks as guess times pi times guilt. Computer science has two hard problems: cache invalidation, naming things, and off-by-one errors. Three items. Accurate. I write undocumented features until QA disagrees in writing. Technical debt is a gym membership I narrate well and visit rarely. Hire me before I refactor this into three services and a webinar.

Experience, approach, and where I come from.

View resume
MigrationWiz for Goog…Building AI Agents wi…AI Agent BuilderBuilding Production-R…Ultimate Prompt Engin…View All

Three ways to work together

Single project

Fixed price
  • One scoped build, web app, AI workflow, or integration
  • Initial contact, then a written plan and a fixed quote
  • Regular check-ins with working pieces
  • Runbook plus free fixes for a follow-up period after launch
Most Popular

Monthly retainer

Most teams choose this
  • Ongoing work as your part-time engineer
  • Set hours per month, rollover for one cycle
  • Slack-level responsiveness during your working day
  • Cancel anytime with 14 days notice, no minimum term

Quick audit

Flat fee
  • A focused deep-dive on your existing app or workflow
  • Written report ranking the top issues by impact and effort
  • Specific fixes you can hand to your team, or to me
  • No build commitment afterwards

Honest answers to questions I actually get

One thing you still do manually every day?

Describe your friction: scattered apps, store ops, inbox chaos, whatever eats your time. I'll reply with whether it's worth automating and what a personalized workflow would look like. No commitment, no pitch.

See my work