FUCHS IT-CONSULTING01 / AI adoption & development
Local models · Cloud models · Agentic development
From a good idea to the right AI workflow.
I support AI projects in highly regulated public-sector environments with high information-security and confidentiality requirements, from needs assessment and model selection through architecture and integration to quality assurance, rollout and training.
Discuss your AI project ↗My approach
I connect the capabilities of modern AI with your tasks, data and workflows, from the first idea to everyday use within your team.
Practice / Project experience
Introduced OpenCode with local models in a development environment with more than 90 developers. The work covered needs assessment and model selection, pilots and integration, usage guardrails, controlled software deployment and training. The client environment is described in generic terms only to protect confidentiality.
Since 2026, I have built numerous software solutions using AI-assisted development workflows, including web applications, analysis tools, interface integrations and automation. I can design use cases, develop the solutions and integrate them into existing systems.
01 / Needs & context
Where could AI support your teams, help access information or accelerate development? Together we identify suitable use cases and translate your goals into a practical starting point.
I bring business teams, IT operations, information security and data protection together, connecting technical possibilities with organisational requirements. The result is prioritised use cases, quality criteria and an agreed path from pilot to adoption.
02 / Consulting areas
Identify needs, prioritise tasks and select suitable pilot cases. Assess value and output quality against concrete tasks and agreed criteria.
Depending on cloud approval, assess OpenAI GPT and Anthropic Claude models or local models. Quality, costs, licence terms and data processing guide the choice. Specific locally deployed models and client operating details are not published.
Connect AI systems, APIs and existing applications. Document interfaces, data access and operational boundaries; design how domain knowledge is incorporated.
Define permitted data, roles, tool access and approval paths. Agree logging, retention and the handling of erroneous or malicious inputs.
Structure software development with Claude Code and OpenCode by combining prompt engineering, agentic workflows, limited access rights, tests and mandatory reviews.
Review results against agreed test cases. Document business accuracy and failure patterns; use the findings to refine models and prompts.
Translate pilots into a controlled introduction. Coordinate provisioning, updates, ownership and fallback options with operations and security teams.
Enable business and development teams for their tasks. Teach effective prompts, critical output review, safe data use and clear escalation paths through practical exercises.
03 / Model strategy
I compare models and tools across providers. Your tasks, expected output quality and whether cloud use is an approved option for the intended data guide the choice.
If cloud is an option: OpenAI GPT models and Anthropic Claude are among the possibilities considered. I support the assessment of capability, costs and deployment, aligning data processing, contractual requirements and integration with your organisation.
If cloud is not an option: Local models enable processing within your own environment. I support model selection and the design of compute resources, provisioning, updates and access protection.
For agentic development: Tools such as Claude Code and OpenCode help structure development workflows. I align tool selection, model connections and data access with your working practices and approved use cases.
For hybrid approaches: We map tasks and data to suitable local or cloud services. Understandable data flows and clear operational ownership provide a foundation for focused improvement.
04 / Development & processes
I develop software using agentic AI workflows and support their introduction within teams. With Claude Code and OpenCode, I connect prompt engineering, implementation, tests and code reviews in a structured development process.
My development experience includes Java, JavaScript and TypeScript, Quarkus, Node.js, React, PrimeFaces and REST APIs. In a highly regulated environment, I am developing a regulatory-reporting monitoring application using Java and Quarkus with a PrimeFaces user interface. Prompt engineering supports development; architecture, tests and business validation remain separate responsibilities.
Integration with BPM, RPA and Intelligent Automation starts with the actual process. Together we design AI-supported steps with understandable results, human review points and appropriate approval paths.
05 / Collaboration
Depending on the engagement, deliverables include prioritised use cases, model assessments, technical architecture and interface documentation, an evaluation concept, guardrails, rollout plans and training materials. Scope and approval processes are agreed for each project.
Profile & AI training ↗06 / Contact
Planning an AI introduction, comparing local and cloud models or looking to establish structured agentic development workflows?
[email protected]Tell me briefly which task you want to tackle with AI. We will agree a suitable starting point.