CV

Automating Map-Making through Enhanced Geographic Information Extraction Using Retrieval-Augmented Generation with Open-source Large Language Models

Currently Drafting, 2025

Comparative Analysis of BERT and GPT for Classifying Conflict News with Sudan Conflict as an Example

Preparing to Submit, 2025

Optimizing context-based location extraction by tuning open-source LLMs with RAG

IJDE International Journal of Digital Earth, Under Review, 2025

• Developed and evaluated ML pipelines using BERT fine-tuning and GPT with RAG, ICL, and hyperparameter tuning to classify conflict incidents, achieving 91% accuracy in single-label classification.
• Extracted conflict locations from news articles using RAG, CoT, and few-shot LLM prompting, achieving > 90% F1 in context-aware information retrieval.
• Develop end-to-end pipelines integrating large language models, QGIS, Neo4j, and Streamlit for automated conflict mapping and analysis.
• Installed and maintained Linux-based operating systems (Red Hat 8, Rocky Linux, Ubuntu) across 7 rack-mounted server units to support high-availability computational workflows.
• Designed and implemented a 3D data center digital twin using NVIDIA Omniverse, visualizing real-time server metrics for intuitive infrastructure monitoring.

Developing real-time monitoring and visualization systems using Prometheus and Grafana for comprehensive data center performance analysis.

Automating conflict mapping via multi-agent system using LLMs, Neo4j, and PyQGIS.

Contributed to machine learning pipelines for classification using BERT, GPT, FAISS, and OpenAI APIs.

Researched context-specific location extraction for conflict articles using LLMs and Retrieval-Augmented Generation.

BlueTemp

AI platform supporting marine environments by predicting sea water temperatures. From forecasting hurricanes to saving coral reefs.

Crushor

A retro-style platform-based game with wall-jumping, dashing, and stomping mechanics.