Emmanuel Osabutey

About Emmanuel Osabutey

I am a Machine Learning Engineer with expertise in designing, optimizing, and deploying scalable AI solutions that bridge the gap between research and real-world applications. My passion lies in transforming complex data into production-ready models and pipelines that deliver measurable impact for organizations and end-users. I specialize in building advanced ML systems using LSTM, BERT, Random Forest, XGBoost, ARIMA, and SARIMA, and I leverage techniques like hyperparameter optimization (GridSearchCV, RandomSearch) to maximize performance. Beyond training models, I focus on deployment and accessibility—ensuring solutions are robust, scalable, and easily consumable by stakeholders through tools like Streamlit, Docker, and Google Cloud Platform (GCP). Some highlights of my work include: • Developing real-time speech translation models (LSTM, BERT) that reduced latency by 20%. • Deploying predictive QoS models to improve network reliability, cutting call drops by 15%. • Building scalable prediction pipelines for healthcare (sepsis detection) and telecom (churn reduction), integrating FastAPI and GCP for production deployment. • Creating monitoring dashboards (Power BI, Tableau, Streamlit) to track model performance and improve stakeholder decision-making. With strong proficiency in Python, SQL, MongoDB, and GCP cloud deployment, I thrive in the end-to-end ML lifecycle—from data engineering and feature selection to model optimization, deployment, and monitoring. I’m passionate about delivering production-grade AI systems that drive real-world value, improve decision-making, and scale seamlessly. Let’s connect if you’re looking to collaborate on building and deploying impactful ML solutions.

Experience

1. Machine Learning Intern

   Vosyn

   Jun 2025 — Present

   Roles and Responsibilities• Assisted in building a speaker diarization workflow (Faster Whisper, NeMo MSDD, and initial T5 cleaning), contributing to a 25–40% reduction in transcript generation time per audio file and enabling near real-time feedback in an internal prototype.• Supported deployment of a FastAPI service and GCS integration, enabling a 2x throughput increase for batch diarization tasks and reducing per-request latency by 30–50 ms on average.• Contributed to data preprocessing and transcription cleaning (NLTK and early T5 experiments), resulting in a 15–25% improvement in readability scores and a measurable reduction in punctuation/grammar errors.

Education

• Azubi Africa

  Certification

• Fitchburg State University

  Msc, Computer Science

• University of South Dakota

  Master's degree, Computer Science

• Kwame Nkrumah University of Science and Technology, Kumasi

  Bsc Telecommunication Engineering

  2017 — 2021