Ribka Tiruneh User

At NEOALI, I developed and deployed NLP systems, including Llama 3 fine-tuning and retrieval-augmented generation workflows for document classification and information extraction over messy production data.

At George Washington University, I reduced a 90,000-node physics-based simulation to 177 nodes while preserving functional behavior. I built the evaluation framework, ran systematic ablations, achieved 91% peak reconstruction accuracy, and created reproducible Python optimization pipelines.

My work spans machine learning modeling, time-series signals, NLP, computer vision, data pipelines, and production-ready APIs. I bring rigorous baselines, cross-validation, uncertainty analysis, documentation, and reproducibility controls to ambiguous research problems.

I've also built predictive decision-support prototypes, published research as a co-author, and defended a thesis on reduced-scale CA3-inspired neural networks.