At KlearNow.AI, I architected and deployed an AGNO-based multi-agent Document AI system using Claude 4.5 Sonnet. It orchestrates six microservices for document classification, OCR, and structured data extraction, processing 3,000+ documents per day with 99% reliability.
I modernized document extraction by moving from LayoutLM and BERT to an LLM-powered architecture that adapts to 1,000+ document templates without template-specific retraining. I also developed a Validation Agent that enforces schema rules, reducing downstream errors and manual correction effort.
As a Machine Learning Engineer Intern at KlearNow.AI, I fine-tuned Qwen and LLaMA with LoRA/QLoRA and optimized prompting strategies. At Beans.ai, I developed a CNN-based application to detect parking, roads, and building entrances, and created indoor maps using ArcGIS.
My projects include email intent classification and routing, multimodal document classification with vector retrieval, and document information extraction. I’ve worked across applied machine learning, LLMs, computer vision, and production ML systems.
