Workflow
How I move from an idea to a working application.
I do not start with code. First, I seek to understand the process: who interacts with the system, what user roles exist, what data flows through the application, where decisions are made, what must be persisted, what can be automated, and what should never be promised. Only then do I design the application structure.
Typical workflow
Problem identification
I define which organizational process requires structure and what currently fails: scattered spreadsheets, message threads, informal verbal agreements, missing change logs, absent statuses, lack of reporting.
Process model
I map out user roles, procedural steps, input data, decision branches, status transitions, exceptions, and boundaries.
Application model
I design screen hierarchies, database schemas, permission matrices, administrative interfaces, and core user journeys.
Prototype and implementation
I build a functional application, typically using a modern web stack. Depending on the project, this includes analytical modules, local client-side data processing, structured exports, or automations.
Documentation, quality control, and demo
I document project status, operational constraints, sample demonstration data, usage guides, and potential expansion paths. In projects requiring formal verification, I also establish audit trails, evaluation criteria, evidentiary records, or quality gates.
How I use AI tools
AI is part of the engineering toolkit, not a substitute for understanding.
I leverage AI tools primarily as an extension of my engineering and software design workflow. They assist in problem analysis, exploring alternative formulations, identifying edge cases and gaps in assumptions, architectural planning, data schema design, implementation and refactoring, drafting documentation, and testing.
What matters most to me is that AI facilitates rapid iteration: from the big picture to granular details and back. Before planning or writing code, I spend substantial time examining the problem domain — both systemic principles and practical edge cases — and repeat these iterations continuously at every stage. Effective use of AI begins well before coding.
What I don't do
- I do not present a project as a production-ready product if it is an exploratory demonstrator.
- I do not promise legal compliance without explicit regulatory baselines, clear assessment criteria, evidentiary proof, and expert validation where required.
- I do not publish confidential or production data as demo data.
- I do not pretend that an AI tool replaces authentic domain understanding.
- I do not treat a ranking, report, or algorithmic score as a decision without human interpretation.