• Led development of a scalable multimodal RAG pipeline for retrieving information from large-scale text and visual datasets, integrating a ColPali-based vision-language retriever.
• Optimized retrieval performance with 4-bit QLoRA fine-tuning, novel 32× patch-reducing pooling, and re-ranking achieving ~93% NDCG@10—resulting in a 10× speedup over baseline.
• Designed a kernel-based embedding fusion method and SmolDocling-powered document parser to enhance LLM response grounding; collaborated with domain experts to productionize chatbot behavior via prompt tuning and FastAPI integration.

Courses:

• Machine Learning
• Natural Language Processing
• Neural Networks and Deep Learning
• Computer Vision
• Big Data and Cloud-Based Data Analytics
• Perception in Robotics
• Reinforcement Learning
• Development of UAVs
• Motion Planning
• State Estimation