Data analyst to build multi-source transaction reconciliation pipeline (Python/AI-assisted)

WHAT WE NEED We are a US-based fintech (LootRush.com) processing high monthly volume across multiple payment rails. We need someone to build a reconciliation pipeline that matches transaction data across three independent sources: settlement files from payment partners, our internal transaction records, and public blockchain transaction data. The output is a repeatable, documented matching process that identifies every unmatched or partially matched transaction and explains why it broke. This is not a one-off analysis. It is a tool we re-run every month. WHAT YOU WILL BUILD 1. An ingestion layer that normalizes files from multiple sources with inconsistent formats, column names, timestamp conventions, and rounding behavior. 2. A matching engine handling exact matches, near matches, and one-to-many relationships, where a single record on one side corresponds to multiple records on another. 3. An exception report that categorizes each break by type and gives a human enough context to resolve it. 4. Documentation clear enough that someone else on our team can run, modify, and extend the pipeline without you. WHAT WE ARE LOOKING FOR - Strong Python and pandas, or an equivalent stack. SQL a plus. - Real experience reconciling or matching messy data across systems that were never designed to agree with each other. Financial data preferred but not required. If you have done this with logistics, telecom, ad tech, or any other multi-source operational data, that transfers. - Fluency with AI tools, specifically Claude, as part of how you actually work. We use them heavily and want someone who does too. We will ask about this concretely. - The instinct to chase a discrepancy down rather than write it off. The unmatched 0.3 percent is the entire point of the job. - Clear written English. You will be explaining your logic, not just shipping code. NICE TO HAVE, GENUINELY OPTIONAL Familiarity with blockchain data and block explorers, or prior work with payment settlement files, interchange, or card transaction data. We will teach the domain. HOW WE ARE RUNNING THIS We start with a paid fixed-price test using synthetic data that mirrors our real structure. It is a realistic file set with a small number of deliberate discrepancies buried in it. You deliver the reconciliation output plus a short writeup of what you found and how you approached it. We pay for the test regardless of outcome. All development happens against synthetic data. We run the finished pipeline against live data internally. Strong performers move to an ongoing engagement. We have continuing work in this area. TO APPLY Skip the generic cover letter. Tell us about one time you reconciled or matched data across sources that disagreed, what the root cause turned out to be, and how you found it.