MSc Data Science graduate focused on applied machine learning, reproducible analytics, and product-minded ML systems.
I build machine learning projects that connect modelling with real-world decision-making: validation design, calibration, explainability, threshold trade-offs, product constraints, and clear communication of results.
I recently completed an MSc in Data Science with Distinction from the University of Aberdeen and am now targeting Machine Learning Engineer, Data Scientist, and Applied AI roles in the UK.
Founder-built Android + ML product case study for a privacy-conscious digital wellbeing app.
GameBrek is a private-source Android product designed to help users create calmer boundaries around gaming, social media, and distraction-heavy apps. The public repository documents the product strategy, UX thinking, privacy-first architecture, ML recommendation design, and Play Store readiness work without exposing the production source code.
Highlights:
- Designed a recovery-first focus product around self-directed app boundaries.
- Built the product story around privacy, behavioural patterns, and realistic user friction.
- Developed an on-device ML recommendation direction to support session planning.
- Prepared the product for public-facing case-study documentation while keeping source code private.
Technologies and skills: Kotlin, Jetpack Compose, Firebase, Android product development, on-device ML recommendation design, privacy-first architecture, UX/product thinking.
Retrospective sepsis early-warning ML analysis using patient-grouped validation and decision-aware evaluation.
This project studies early sepsis risk modelling from patient time-series data. It focuses not only on predictive performance, but also on calibration, threshold selection, alert burden, and lead-time trade-offs — the kinds of issues that matter when ML outputs are connected to real operational decisions.
Highlights:
- Built temporal features from patient-level time-series data.
- Used patient-grouped validation to reduce patient-level leakage risk.
- Compared model performance using saved cross-validation artifacts.
- Analysed calibration, threshold policies, alert burden, and lead time.
- Communicated healthcare ML limitations carefully as retrospective analysis, not clinical deployment.
Technologies and skills: Python, pandas, scikit-learn, temporal ML, grouped validation, calibration, threshold analysis, healthcare ML evaluation, reproducible analytics.
End-to-end churn prediction project focused on practical business value and explainable model outputs.
This project applies supervised machine learning to customer churn prediction, with emphasis on understanding business drivers, comparing models, interpreting predictions, and translating results into retention-focused recommendations.
Highlights:
- Performed exploratory analysis to understand churn patterns.
- Built and compared classification models.
- Used explainability to interpret important churn signals.
- Connected model outputs to business-facing retention insights.
- Structured the project for portfolio-ready communication.
Technologies and skills: Python, pandas, scikit-learn, classification, EDA, model evaluation, SHAP/model explainability, business interpretation.
Computational design of RNA thermoswitches for high-temperature genetic control in Bacillus subtilis.
This MSc research project explores computational design approaches for RNA thermoswitches, combining sequence design, thermodynamic modelling, and machine learning-style evaluation to support biological design decisions.
Highlights:
- Worked on a scientific ML/computational biology problem.
- Used computational tools to support RNA sequence design.
- Evaluated sequence behaviour through thermodynamic and feature-based analysis.
- Communicated a technical research workflow as part of an MSc Data Science project.
- Demonstrated domain flexibility beyond standard business datasets.
Technologies and skills: Python, NUPACK, pandas, scikit-learn, sequence design, computational biology, feature engineering, model evaluation, research communication.
- Applied machine learning: classification, model comparison, feature engineering, calibration, threshold analysis, and explainability.
- Reproducible analytics: clear project structure, documented workflows, validation-aware evaluation, and result interpretation.
- Product-minded ML: connecting model behaviour with user experience, privacy, operational constraints, and decision-making.
- Healthcare and scientific ML: careful framing of limitations, data leakage risks, validation choices, and responsible communication.
- Core tools: Python, pandas, NumPy, scikit-learn, Matplotlib, SHAP, SQL, Git, GitHub, Jupyter, VS Code, Kotlin, Jetpack Compose, Firebase.
I am currently focused on building ML and data science projects that are:
- Validated with realistic data-splitting and evaluation choices.
- Clear about trade-offs, limitations, and deployment assumptions.
- Designed for reproducibility and readable technical communication.
- Connected to business, product, healthcare, or scientific decision-making.
- LinkedIn: Md Abir Hossain