Step in detail
A step built through practice
Starting from a research need
At the Grenoble Computer Science Laboratory, the goal was not to produce a decorative chart. Responses from a sociological questionnaire about life trajectories had to be represented in a way that was readable, editable, and faithful to the researchers' method. I therefore began by understanding the vocabulary, dimensions, and edge cases of the model being studied.
Building a genuinely interactive visualization
The CAP2vie case study led me to transform exact or estimated periods into a D3.js visualization distributed across five dimensions and available in two reading modes. The form and chart had to remain consistent when an interviewer corrected an answer. The display evolved through iterations rather than added effects until it provided a usable demonstration foundation.
Working with non-technical experts
Discussions with researchers were part of the development: reformulating a request, showing a version, collecting an objection, and translating it into a data or interface change. This experience strengthened my communication and taught me to defend a technical decision without losing the product's scientific meaning.
Keeping data, rules, and interpretation aligned
The CAP2vie record shows that the challenge was also one of consistency: a period could be exact or estimated, an answer remained editable, and its representation had to change without contradicting the questionnaire. I therefore worked at the boundary between the PostgreSQL data model, the Vue.js application, and the D3.js output. This taught me that a visualization is a business feature: its usefulness depends on faithful transformations, clear labels, and the ability to correct a case before interpreting it.
A demonstrable foundation with an explicit scope
The documented outcome is a first functional version usable for demonstration, not a generalized research product whose large-scale use had been measured. This distinction matters: it prevents presenting prototype validation as evidence of scientific impact. The work did, however, establish a reusable way to evolve a tool from expert feedback while preserving traceability between a request, a data rule, and a visible change.
