Serverless Function Patterns

Serverless functions work best when each call has a clear boundary: inputs arrive, work happens, outputs return or land in storage.

Good candidates

• Convert many files in parallel.
• Score many independent batches.
• Run short GPU experiments.
• Generate previews or validation reports.

Boundary sketch

@coiled.function(region="us-west-2")
def convert(uri):
    df = pandas.read_csv(uri)
    out = uri.replace(".csv", ".parquet")
    df.to_parquet(out)
    return out

Design habit

Make retries safe by giving each call deterministic inputs and outputs.