
Michał Kasperczyk
Senior .NET engineer · AI systems for the SDLC
Senior .NET engineer with 15+ years of experience, currently building AI systems that support the SDLC. Six years leading teams; here I write about agent harnesses, verification gates and the architecture behind reliable AI systems.
AI systems for the SDLC
My current focus: building AI systems that support the software delivery lifecycle — support tooling, integrations with the systems teams already use, the libraries worth knowing, and what it takes to run all of it at scale.
The agent harness
I'm fascinated by the harness around coding agents — skills, automation, and everything that makes an application easy to develop well. I test the tooling hands-on: Claude Code, GitHub Copilot, and whatever ships next.
Leadership, pragmatically
As a team leader I built a development department from nearly zero: kept a large production system running and shared knowledge until the teams were fully autonomous — and still shipping new things. Process followed the work, not the other way around: Scrum first, Kanban when the way of working changed, back to Scrum once the teams ran smoothly.
Engineering foundations
DDD, system design, and the long life an application has after the demo. Backend mostly C#/.NET, with reporting systems built on Databricks and Python. The front end has always mattered to me — from jQuery through Angular to React, including introducing a monorepo at Stretto. I've taken applications all the way from planning to production and demos.
Applications playground
The applications behind the write-ups, running live on my own hardware.
-

Aggregate: Where You Draw the Lines (and What It Costs to Draw Them Wrong)
12 min readDDD aggregate design from first principles: the four rules, how to find the consistency boundary, and what it costs when you draw…
-

Five Bounded Contexts, Three Services: A Tour of a Warehouse Domain
10 min readBounded contexts in practice: a tour of five contexts mapped to three services, how they talk through events not shared types, and…
-

Writing Agents That Work: opening a field guide from skills to systems
4 min readA field guide to building agentic systems — skills, subagents, feedback loops — that separates peer-reviewed evidence from vendor self-report. Sources tagged.
-

From PR to Production: Why We’re Building Forge
6 min readPart 1 in the series on building Forge, an agentic SDLC that runs a ticket from a business request all the way…
-

Opening the Forge Series: Design Before Code, Measurement Before Claims
3 min readOpening the Forge series on building an agentic SDLC from ticket to production — design before code, and a rule that every…
-

Why We Start With the Domain, Not With Docker
18 min readWhy domain-driven design starts with the business, not `docker compose up`: the big-picture board, the strategic decisions it surfaced, and the price…
-

Event Storming: Modeling a Domain on a Wall Before You Write Code
9 min readEvent Storming explained: the colour grammar, the three zoom levels, and how a wall of sticky notes models a domain before you…
-

How Microservices Are Really Built: A Series
5 min readI want to show you how microservices are built in real life. Not the conference-talk version, where three perfectly named services exchange…
-

PostgreSQL vs MSSQL Locking: Architectural Differences and Practical Patterns
4 min readMastering Database Concurrency: From MSSQL Lock Hints to PostgreSQL MVCC.
-

When and How to Use Locking Hints in SQL Server?
5 min readIn a perfect world, the SQL Server Query Optimizer works perfectly. It automatically decides when to use Shared (S) locks for reading…