Caleb Leung

About Caleb Leung

Hi there! I am a first-generation recent graduate with a Mathematics bachelor\'s degree…

Experience

1. Software Engineer

   Intellpro

   Oct 2024 — Present · New York, NY, US

   Built a Python contextual-injection pipeline for RAG LLMs that ingests HTML/PDF SEC filings (e.g, 10-K), parses tables and bold headers (regex + CSS selectors), applies spaCy-based chunking, and normalizes content into headers, text chunks, and tables.• Implemented hybrid matching—semantic (SentenceTransformers)+ fuzzy (RapidFuzz)—to align predefined section headers across heterogeneous filings, improving cross-document consistency.• Hardened HTML/CSS extraction and encoding/decoding to prevent data loss and increase parser robustness across document variants.• Wrote pytest coverage for data collection → digestion layers; set up async tests with mock S3 and a mock Postgres database.• Implement a custom chunking library from scratch tailored to financial documents (instead of LangChain), capturing tables intact, avoiding sentence breaks, and reducing information loss during chunking. • Built a classification model to identify the correct category for each financial documents, including data embedding, choosing scoring function, perform training, achieve 98.5% accuracy

Education

• Baruch College

  Bachelor of Science - BS