Automation of Funding Request Data Validation

How do we balance speed and spending surrounding the use of construction funding requests?



During my internship with Intel Corporation in 2023, I was tasked with the automation of a manual review process for construction funding request data. The manual process was often complicated and time variant, due to the complex and imperfect nature of the data. The automation, built in Python, has the potential to reduce the required review time from 3 hours to 15 minutes per funding request, and reduce the amount of review cycles required per record.

The process automation, architected in Python, utilized file scraping to compare values and raw formulas stored in excel for accuracy and consistency with expected values. The automation also handled inconsistencies through implementation of fuzzy matching algorithms and strict data preprocessing. The final deliverable produces a raw data file of anomalistic values to be converted via pre-formatted pivot tables to a usable format for the reviewers and stakeholders. Histori cal logging allowed for a full database of records from prior automation run times and all forward looking requests. This database demonstrated top performers and offenders within the list of contractor partners, as well as what discplinary types were most likely to have an offense. By categorizing types of errors, analytics also demonstrated common anomalistic behavior, and could be used in avoidance tactics moving forward.

In full, by automating the process, data quality increased within the construction team, process time and manual touch requirements were reduced, the project was scoped for extension to other sites/contractors, construction projects moved up on the calendar, and cost avoidance opportunities became abundant.

The following python libraries were integral to the project:

  • nltk
  • pandas
  • numpy
  • pylightxl & openpyxl
  • difflib
  • datetime