I wish the "introduction" section of papers like this one wouldn't immediately use jargon that has me guessing what they're talking about.

What is "organic user activity"? A "lambda model combining batch and streaming paradigms"? Does "persisting messages" mean resending messages? What does "treat inputs as the company's golden source of truth" mean? "Expose the abstraction of independent data streams"? "Mask out heterogeneity"? "Deserialization tax"? "Heterogeneous use cases"?

I can uncomfortably guess what they're trying to say, but I'd much rather they used the vocabulary of a six-year-old to hook me in the introduction.