Experience

Software Engineer Intern

Jiangxi Sincere Mineral Company · Jiangxi, China · Jun 2025 – Sep 2025

• Built a Django REST + PostgreSQL service that replaced manual re-keying of supplier invoices: an LLM extracts 12 key fields (invoice no., date, tonnage, amount) per invoice into a pre-filled review screen, reaching over 95% field-level accuracy on a 30-document hand-labeled test set and cutting per-document entry time by 50%.
• Moved document intake to an asynchronous pipeline — the upload endpoint writes the file to S3, enqueues a Celery task on Redis, and returns a job ID in under 200 ms while the frontend polls a status endpoint — so 30–60 second LLM extractions no longer exceeded the server’s 30-second request timeout on large scanned PDFs.
• Designed an Adapter layer that normalizes scanned PDFs, supplier Excel workbooks, and CSV exports into one 12-field invoice schema; adding a second supplier’s Excel layout later took one 80-line adapter class with no changes to the extraction or validation code.
• Implemented two-tier duplicate detection in the service layer: an exact match on invoice number returns a blocking Error, while a fuzzy match on (supplier, date, tonnage, amount) returns a non-blocking Warning the operator can override — eliminating double-entry into the settlement records without rejecting legitimately reissued documents.
• Extracted duplicate-detection and settlement logic out of Django views into a service layer behind repository interfaces, enabling 40+ pytest unit tests against mocked repositories and a stubbed LLM client that cover 90% of the service layer and run in 3 seconds with zero API cost.

Carnegie Mellon University

Master of Information Systems Management · Pittsburgh, PA · Aug 2026 – Dec 2027

Coursework: Large Language Model Systems, Data Structures and Algorithms.

University of Washington

Bachelor of Arts, Mathematics & Philosophy · Seattle, WA · Sep 2021 – Jun 2025

Programming

Tools