Complete filtering of your own data sources - without losing the path that got you there.
Client
Under NDA
Disciplines
Software development, data engineering
Form factor
Web application
Engagement
Production platform under NDA
The client works with large list and operational files. Filtering in Excel overwrites the last step. Nobody can say, three days later, exactly which conditions produced the file that went out.
They needed complete filtering - nested conditions, reusable sequences, exports - with history that behaves like a tree, not like undo.
The hard part is not drawing a filter UI. It is keeping every derived set cheap, correct and explainable.
The model is parent dataset plus transformation equals derived dataset. Root has no filter. Every other node applies a keep or remove, optionally scoped, and caches the matching row ids. Rows are not duplicated per branch.
Saved transformations and saved sequences are definitions, not pointers at live nodes, so a template can be applied to a new import without dragging old data with it.
A three-pane web application: source tree, filter builder, and a virtualized table of the current node. Import wizards cover delimited files and fixed-width layouts. Any node can export the result, the lineage, or both.
The same discipline - do not destroy the last good view - is what we apply when a team later asks for workflow automation or a web scraper feeding this kind of source.
The engagement is under NDA. We do not publish the client's lists, volumes, or commercial outcomes.
The platform is used on the client's own sources. New imports attach to the same tree model; new filter sequences are saved and reused rather than rebuilt.
This engagement is covered by a confidentiality agreement, so the client is not named and some implementation detail is described in general terms.
Next project
SV Connect Fleet Management