Emmanuella Ewurabena Turkson

I’m a Computer Science student with hands-on experience building data-driven systems across machine learning, LLM evaluation, and full-stack product work. I’m energized by turning messy inputs into measurable performance—then communicating results in a way teams can act on.

At Dell Technologies, I built an automated multi-model evaluation pipeline in Python and LangChain benchmarking 3 LLMs for accuracy and error detection, cutting manual security triage time by 40%. I also engineered a CLI-based risk detection tool for command-injection and root-shell patterns, and fine-tuned/evaluated 4 large language models for vulnerability detection with a 6/6 detection rate.

At LBNL, I engineered a reproducible ML pipeline with full unit test coverage for selectivity prediction on 50K+ molecular records, accelerating iteration and peer review. I improved validation accuracy by 15% using transfer learning and data augmentation with CNNs, and I collaborated with multidisciplinary scientists and engineers to shape next-phase generative modeling strategy.

I also bring product and research discipline from my internship at the CFPB, where I analyzed 200+ survey responses and 5 focus groups to surface core pain points for policy recommendations. I designed and user-tested the CreditBoost prototype, improving targeting and increasing student outreach by 25%, and I continue to value clarity, safety-minded UX, and rigorous evaluation in every build.