Semantic Kernel, not slideware.
LLM and agent systems that survive contact with production data, real users, and the day the model is confidently wrong.
Semantic Kernel · RAG · Agents
Ronneby · Established 1999
A small team with
A small team with
disproportionate
firepower.
Noisy Cricket is a two-person workshop in Ronneby, Sweden. Twenty-five years of shipping AI, .NET, and Blazor systems for clients who prefer a sharp scalpel over a larger team.
25+
years of shipping, same two hands
Trusted by leading organisations
How we work.
Five services, one team. We engage deeply on each, often for years at a time. No account-manager layer, no junior triage, no handoff latency.
Blazor systems that keep shipping.
.NET 8 web applications, APIs, and service platforms — built for a ten-year lifespan, not a launch-party demo.
Blazor · ASP.NET · AzureDrev — our own storefront engine.
Runs on a single server without load balancing and still outperforms vendor platforms costing ten times more.
Drev · Custom · High performanceLearnways — we write the backend.
Backend engineering and Azure operations for Learnways' portfolio of web-based course sites — .NET Core, SQL Server or SQLite, our Azure footprint, since 2003.
Learnways · .NET Core · Azure · BackendMPS that runs the line.
Production systems controlling industrial processes to automotive-standard tolerances. Twenty-four-hour uptime.
MPS · Industrial · 24/7You email the people who write the code.
No third layer. No Slack triangle, no escalation path. Johan or David, straight from the keyboard.
Two humans · Always25+
Years in business
90+
Client engagements delivered
2
Senior developers, no juniors
8y
Average engagement length
Selected work.
Eight engagements from the past decade. The full clients list spans Volvo, Pfizer, the United Nations, and roughly eighty more over twenty-five years.
How we think about AI.
Most "AI integrations" in the wild are a large language model bolted onto a product roadmap as decoration. That is not integration. That is signage.
Start with the boring question.
What decision, made by a human today, can a model make better, faster, or at a scale your team cannot sustain?
Data in · data out · failure modePlan for the day the model lies.
Every Semantic Kernel we ship comes with escalation logic, fallback paths, and an audit trail a human can read.
Guardrails · Citations · AuditOutcomes, measured in money.
No demo is finished until we can point at the line in the P&L where the integration shows up.
P&L · SLA · ROIGot something loud to build?
We take on a small number of engagements each year. Tell us what you are shipping — we will tell you honestly if we are the right team for it.
Write to us