biye1

Designed and engineered a full-stack ETL internal tool that automated e-invoice data extraction and templating on customer-level datasets, boosting extraction speed by 11% through an LLM. Built backend with Python and Flask, and frontend with TypeScript and React. Fine-tuned a language model using Python to achieve over 90% accuracy in detecting formatting errors in financial data. Developed a gradient-based classifier for tax-report data, achieving an 85% accuracy rate in identifying potential tax fraud. Partnered with developers to maintain a SaaS tool built with Azure Databricks. Utilised Agile methodologies and Scrum principles to ensure timely delivery and alignment with business requirements. Advised non-technical stakeholders on AI strategies, automating time-consuming tax processes & streamlining workflow.

Partnered with five external engineers to develop a scholarship donation platform, using GitHub for version control: Next.js and TypeScript for the frontend, and Supabase for the backend SQL database. Successfully delivered the project by the summer of 2024. Conducted technical meetings with external engineers, aligning app functionality with stakeholder needs and feedback, while meeting with non-technical stakeholders to encourage communication. Platformed initiatives providing funding opportunities for students under the mentorship of a member of the Board of Trustees at Meta, leading to the launch of 7 new scholarships totaling over £50,000.