Mark Asare

Developed and migrated a predictive order completion system from AWS to Azure, deployed to AKS with Azure ML training pipelines and Azure DevOps CI/CD, reducing delivery times; provisioned ML infrastructure with Terraform and built Synapse-Snowflake data pipelines for production ML.

Transformed five Excel decision tools into web-based Flask applications and developed an end-to-end chemical usage forecasting system with TensorFlow, plus predictive maintenance models that reduced downtime and maintenance costs.

Built and productionised time-series forecasting models for renewable energy output and automated data ingestion and reporting pipelines to deliver reliable daily Power BI updates for operational planning.

Established automated testing and validation frameworks for deployed ML models, reducing production error rates, and supported deployment of large-scale document processing transformer models achieving >95% extraction accuracy.

Led development of NLP classification and conversational AI models using TensorFlow and PyTorch and delivered data processing and text analytics pipelines leveraging spaCy, NLTK, and custom preprocessing.

Trained and optimized deep learning models including Tacotron2 and NEMO for speech tasks and developed sentiment analysis on speech-to-text outputs using BERT and XGBoost to improve customer satisfaction analytics.

Developed R-based predictive models using caret and glmnet to estimate loan default probabilities, improving risk assessment accuracy for financial clients.

Completed an MSc in Stochastics and Data Science focusing on probabilistic modelling and data-driven methods for machine learning applications.

Earned a Bachelor of Arts in Economics and Statistics with coursework in econometrics, statistical modelling, and data analysis.