Alternative data pipeline
[One sentence: what the pipeline ingests, who the analysis was for, and the decision it was meant to inform.]
The problem
[Two paragraphs on the question you were trying to answer, and why the obvious data source could not answer it. Alternative data only earns its cost when the standard feed has a blind spot — name that blind spot precisely.]
[What made it hard: rate limits, schema drift, licensing, sparse coverage, survivorship bias in the panel. This paragraph is what separates a portfolio from a résumé bullet.]
What I built
[Vendor feed]
[Filings]
[Retry & backoff]
[Schema checks]
[Normalised tables]
[Signal testing]
[Coverage & bias checks]
[Reporting]
[Walk through the decision that mattered. Why you stored raw and immutable before normalising. How you handled a source changing its schema without warning. What you deliberately left out of v1, and why that was the right call.]
This section is plain markdown, so lists, links and code blocks all work:
# A short, real snippet beats a paragraph describing one.Results
What I'd do differently
[One honest paragraph. Interviewers ask this exact question, and having the answer already written is a real advantage. It also signals seniority better than any metric on this page.]