We would like to use Google Analytics to understand how our website is used and to improve it technically and editorially. Google Analytics is only loaded after you give your consent.
From the initial idea to an operational AI solution: software architecture, machine learning, language processing and language models in one system.
Technologies:
Python · spaCy · Natural Language Processing · Machine Learning · OCR · Large Language Models · Local LLM
[ 03 // POSITIONING ]
Why EicDevCon?
Technical projects need more than working code. What counts are solutions that fit into existing structures, stay understandable, and remain manageable over the long term.
Experience in demanding IT environments
Software development and integration that takes existing systems, technical dependencies, security and dependable operation into account.
Technical understanding beyond implementation
Architecture, interfaces, data flows, operations and data protection are considered together — rather than each requirement being built in isolation.
Direct and transparent collaboration
Short lines of communication, requirements assessed on their technical merits, and decisions that stay traceable without unnecessary organisational layers.
Sustainable, maintainable solutions
The focus is on maintainable systems, architectures that can be reasoned about, and keeping technical dependencies to a minimum.
EicDevCon works with organisations where standard solutions fall short and technical decisions carry consequences well beyond the project.
[ 04 // ENGINEERING PROOF ]
The stack is a means, not a statement.
Technologies appear here in the context in which they are actually used.
Application engineering
Websites, business applications, interfaces and data flows – where technical requirements, domain logic and existing systems come together.
PHPServer-side applications and interfaces, including this website.
PythonData processing, automation and integration scripting.
TypeScriptType-safe application logic in the browser and in Node tooling.
PostgreSQLRelational storage with dependable consistency guarantees.
Mobile
Native applications for Android and iOS, including on-device data handling and release processes.
KotlinAndroid applications including on-device storage.
SwiftiOS applications following platform-native interaction patterns.
AndroidPlatform-specific requirements, permissions and store processes.
iOSReview requirements, privacy declarations and release cycles.
Offline-firstApplications that remain fully usable without a network connection.
Identity & access
Authentication, authorisation and user management across systems that were never designed to interoperate.
KeycloakCentral authentication, realms, clients and role models.
OpenID ConnectStandardised sign-in for web and mobile clients.
OAuth 2.0Delegated authorisation between services.
SAMLIntegration with existing enterprise identity providers.
LDAPDirectory services as the system of record for users and groups.
SCIMAutomated provisioning and deprovisioning of user accounts.
AI engineering & applied AI
Machine learning, language processing and language models as part of a business application – not as a demonstration beside it.
spaCyLanguage analysis and entity recognition in unstructured text.
Natural Language ProcessingMaking linguistic structure usable where fixed rules would be too rigid.
Machine LearningTrainable classification where patterns cannot be enumerated exhaustively.
OCRText recognition in scans when a document carries no usable text layer.
Large Language ModelsSemantic understanding where rules and classification reach their limits.
Local LLMRunning language models inside your own infrastructure, for example via Ollama or LM Studio.
Platform & operations
Runtime, operations and availability – the layer where architectural decisions show their cost first.
LinuxOperating system foundation for application and database services.
DockerReproducible runtime environments for development and operations.
KubernetesOrchestration where several services are operated together.
HochverfügbarkeitFailure behaviour, redundancy and operational boundaries – settled before the incident.
Privacy by DesignData minimisation as an architectural decision, not a later setting.
A document passes through nine consecutive stages: PDF or scan, text recognition, rule-based analysis, language processing, entity recognition, classification, language model analysis, domain logic and structured result. The result is then handed to the business application through an interface.