Ezekiel Jagle

At CVC Pharma LLP, I architected URS Bot, an enterprise document-intelligence platform combining rule-based automation, RAG, and LLMs that improved document-processing efficiency by 90%.

I fine-tuned Sentence-Transformers MiniLM and Nous Hermes models for LangChain and FAISS RAG pipelines, reducing manual document analysis by 70%. I also built feedback-driven data curation and retraining workflows that increased production NLP accuracy by an additional 20%.

I design evaluation and A/B testing frameworks for NLP models, build semantic search services over Microsoft SQL Server, and deploy modular Flask REST microservices with Docker. My document extraction work combines LayoutLMv3, PaddleOCR, and Table Transformer.

Previously, at Siddhi Trading, I built a real-time automated trading bot and high-frequency financial-data ingestion pipelines with 99% data integrity. I also build Python projects including housing-price prediction models and a voice assistant with speech-to-text and text-to-speech.