Hire Dylan S. - Machine Learning Engineer

I worked on the Watson NLP team as one of the project's core maintainers. I made contributions to IBM's chief embeddable natural language processing library, trained and evaluated new NLP models, engaged with external customers, and fostered collaboration among the contributor community. Some highlights of my time in this role include: - Developed and implemented tone and emotion text classification algorithms, resulting in 20-30% improvements in model quality vs legacy cloud service - Engineered a pipeline for fine-tuning foundation models on downstream NLP tasks, enabling automated report generation and reducing manual effort considerably - Implemented containerized model evaluation framework in Kubernetes, reducing model deployment time by 50% compared to legacy monolithic framework - Served as team scrum master for 9 months, orchestrating task assignments, facilitating regular planning sessions, and communicating progress / bottlenecks to higher management Throughout my time as a Machine Learning Engineer at IBM, some skills and technologies I used included Python, deep learning frameworks (i.e. TensorFlow, PyTorch), Scikit-Learn, SciPy, NumPy, and Kubernetes.

I worked in IBM's flagship, incubator-style internship program as one of 24 technical EB interns in the United States. I collaborated with a team of 2 other technical interns and 1 MBA candidate to create a generalizable workflow intelligence application in Python that uses anomaly detection and machine learning techniques to quantify the holistic health of IT processes (i.e. cloud application deployment). My specific role included: - Streamlining data wrangling and normalization procedures across 20+ disparate logging and telemetry data sources - Engineering large-scale (80+GB per day) probabilistic models to measure component-wise process health - Conducting performance tests of different statistical and machine learning methods that we incorporated in our tool - Communicating with stakeholders and potential internal customers to discover painpoints and receive continuous feedback on our incremental progress throughout the summer At the end of the internship, we presented our work to a panel of IBM senior VPs. We were also successfully able to hand off the proof-of-concept MVP to a team of full-time employees for future development. Our solution lowers troubleshooting times by up to 70% and has potential to reduce IBM's site reliability costs considerably. Our team is currently pursuing multiple patents for novel work in the domain space of workflow mining and intelligence.