Data quality & reporting
Reliable data across a global CRM
Automated Python ETL pipelines combined source extraction, data cleaning and school matching to keep Salesforce affiliation data up to date.
Faria Education Group · October 2023–May 2026
The problem
The previous process involved manually collecting partner affiliation lists and comparing them with Salesforce. Teams needed to distinguish new schools from existing institutions with additional affiliations, including same-name campuses and unrelated schools in different locations.
My contribution
I built pipelines to extract partner data from exports, web sources, PDFs and images, then clean missing, garbled or misaligned values. Fuzzy matching compared school and institution attributes with Salesforce. A locally run LLM through Ollama added binary accept/reject decisions using name, address and other matching evidence. Human review remained part of the workflow before Data Loader imports.
The outcome
The pipelines ran in the background to prepare current affiliation data and import-ready outputs. In partner datasets covering at least 6,000 schools, manual evaluation by me, the Data Solutions Principal and the COO found approximately 97% accuracy, with no false positives observed in the reviewed results. Remaining errors were mainly missed matches involving alternate names, renamed schools or differing addresses.
Let’s talk about your data.
I’m interested in work that connects reliable data, useful reporting and practical automation.
Get in touch