MIDAL✳
FROM QUESTION → DATASET
RESEARCH DATASETS ARE A MESS OF TABS, COPIED VALUES, MISSING CONTEXT AND FORGOTTEN SOURCES.
Natural-language research questions are easy to ask. Building a defensible dataset to answer them takes planning, discovery, source evaluation, extraction, cleaning and review.
START WITH A QUESTION. KEEP THE EVIDENCE.
MIDAL is an AI-powered research-to-dataset platform that turns a question into a structured, editable dataset. The key idea is not just collecting values: it is keeping source, evidence and confidence attached so people can inspect and improve the work.
A SMALL, HONEST DEMO.
This interactive sketch illustrates dataset planning and review. It does not run a real backend or claim that its placeholder rows are research findings.
Ask a question. See how a dataset gets planned. This is a deterministic demo, not live research.
WHAT I BUILT
A functional prototype exploring question intake, blueprint generation, research workflow, structured rows, evidence and review.
WHAT FAILED
A value without provenance is hard to trust. Simply generating rows is not enough.
WHAT CHANGED
Source tracking, confidence, evaluation and human review became part of the product model—not a final checkbox.
WHAT I LEARNED
A useful dataset is not just a table. It is a set of claims someone can inspect.
WHAT’S NEXT
Keep improving the workflow, quality checks, cleaning and export.