Case studies

Completed work showing the kinds of tasks you can entrust to us. Four cases from the initial problem to the solution, covering system renewal, improvements to development processes and manufacturing applications.

Optimizing healthcare software development

AI-assisted diagnosis and fix preparation with developer approval

The challenge

Bug fixing is often slowed down by understanding system interactions and reproducing the issue. The knowledge needed is spread across applications, documents and developers. Before AI support, 2-3 bug tickets were resolved per week.

The bug-fixing workflow

Git + AI Knowledge BaseVersioned system knowledge

JiraBug ticket
AI Agent / LLMRoot Cause Analysis
SandboxDocker / Kubernetes · JUnit · GitLab CI/CD
GitLabMerge Request and developer approval

The diagnosis and test results are added to the Jira ticket.

The solution

The working solution uses an AI knowledge base versioned in Git, linking architecture, dependencies, interfaces and data models to source code. A Jira ticket starts targeted code analysis and isolated issue reproduction.

The AI identifies the root cause and prepares two alternatives. It can implement the prepared fixes, verify them with automated tests and prepare GitLab Merge Requests. This accelerates delivery, while the final decision and approval remain with the developer.

Quick Fix: Minimal changes, aiming for a low risk of regression.
Structural Fix: A broader solution, with refactoring where needed.

Results

The comparison uses the number of bug tickets resolved per week before AI support and during AI-assisted development.

bug tickets resolved per week: 2-3 → 5-8Before → With AI support
Nearly three times the weekly output: 2.6×Using the midpoints of the weekly ranges: 6.5 ÷ 2.5 = 2.6.
AI prepares the fix.: Developers decideThe fix can also be automated if required.

Gradual modernization of public-sector software

More stable operations and renewed technology alongside daily work

The challenge

In a public-sector system with dozens of applications, outdated components, slow queries and security gaps hindered daily work. Modernization had to run alongside ongoing operations and business development. Performance targets were page loads under 4 seconds and response times under 2 seconds.

Main areas of modernization

Outdated components: Standardized Java environment, Tomcat and dependencies
Slow queries: Faster data handling through indexing and SQL optimization
Build, test, release: Automated checks and releases with CI/CD

Jenkins, GitLab and SonarQube in the development workflow.

The solution

The assessment prioritized critical defects, transaction handling and database bottlenecks. Further steps were scheduled based on risk, effort and return.

Instead of a complete rewrite, we used gradual refactoring, replaced the Java environment and application server, and standardized dependencies. Modernization included a White Box security review and the introduction of CI/CD.

My role: Architecture design, hands-on development and technical leadership of the team.
Implementation: Scheduled migration, automated tests and documented releases.

Results

Memory use and response times were examined through profiling. AWR/ASH reports were used to analyze database load and queries.

lower memory use: 30-40%After profiling, refactoring and optimization
faster complex queries: 40-60%Through indexing and SQL optimization
page load time: 3-4 secInteraction response times fell to 2 seconds.

These performance figures were documented during modernization and apply to the applications and queries concerned.

Assessment of manufacturing data flows

Reducing repeated data entry and information loss at a manufacturing company

The challenge

Production preparation relied on several separate applications and Excel spreadsheets. The same data had to be entered repeatedly, while changes and program versions were difficult to track. Much of the operational knowledge was held by a few colleagues.

The proposed data flow

Design and ERP dataData from existing systems

Shared data managementImport, validation and versioning
WorkflowStatuses, owners and approvals
Production preparationProgram and document management

The proposed solution

In the short term, extending data transfers, standardizing identifiers and validating input could improve the workflow. The proposal also covers documenting operating rules and improving access security.

The longer-term plan uses shared, versioned data management and a workflow with traceable statuses, owners and approvals. Gradual rollout would be aligned with the existing systems.

Methods: Analysis of system connections and data flows, documenting manual steps and error points.
Implementation plan: Quick improvements and longer-term integration, introduced gradually.

Expected results

less manual data entry: AutomationImporting design and ERP data could reduce typing errors.
recorded statuses and owners: TraceabilityChange histories and approvals would be retrievable.
lower key-person risk: Knowledge transferDocumented operating rules could help new colleagues get up to speed.

The assessment and improvement proposal are complete. The operational benefits shown depend on implementing the recommendations.

Web-based control for automotive manufacturing

Configuration, process monitoring and quality reports for a Poka-Yoke system

The challenge

Development of an automotive manufacturer's web interface had stalled. Understanding the system and connecting the interfaces took longer than planned. We took over the work while retaining the original delivery deadline.

The web interface within the system

Web-HMIBrowser dashboard connected to the virtual application server.

Oracle DB: Central data exchange and control from the web interface through database records
IBM MQ: Message-queue data exchange connected to the Oracle database
Siemens PLC: Production-line control linked to the Oracle database

The solution

We assessed the system and its connections on site. Within one week, a configuration interface for essential settings was ready. Over the following months, we extended it to the full agreed functionality.

The browser dashboard provided process control, configuration, fault tracking, queries and measurement reports. Control operated through the Oracle database, which also connected IBM MQ and Siemens PLCs.

Method: On-site system assessment, prioritizing essential functions and gradual expansion.
System environment: Virtual server, Oracle DB, event-driven IBM MQ data exchange and Ethernet-based PLC communication.

Results

The durations start when we took over web interface development. The first usable version was completed in the first week of the four-month development period.

to the first usable interface: 1 weekEssential browser-based settings during the on-site work.
total development time: 4 monthsProcess control, fault tracking, queries and measurement reports.
After taking over the stalled project.: On timeWith the full agreed scope, connected to the existing system.

The assignment covered developing the web interface and connecting it to the existing system.

How we work
together.

Understanding the situation

Bottlenecks, system connections

Solution plan

Options, priorities, schedule

Implementation

Development, verification, documentation

Facing a
similar task?

Has development stalled, does your system need renewal, or is there too much manual data entry? Not sure where incorrect data enters the system, or already know the cause and want to prevent manual errors? Let's discuss a solution.

The first 30-minute conversation is without obligation. We arrange a time by email.

Let's talk contact@core-systems.hu

Case studies

The challenge

The solution

Results

The challenge

The solution

Results

The challenge

The proposed solution

Expected results

The challenge

The solution

Results

How we worktogether.

Understanding the situation

Solution plan

Implementation

Facing a similar task?

How we work
together.

Facing a
similar task?