Nana Safo Duker

AI/ML Bioinformatics & Precision Medicine - Integrated, Multi-Omics Project Portfolio

A curated multi-project lab of reproducible AI/ML and statistical genomics workflows. Each project is a complete pipeline focused on a specific precision medicine problem—from variant pathogenicity prediction and transcription-factor binding modeling to gene expression inference and biomarker discovery.

This portfolio integrates:

• Deep learning architectures (CNNs, RNNs, Transformers) for genomic sequence analysis
• Statistical genetics and biostatistical modeling for population-level insights
• Explainable machine learning to interpret biological mechanisms
• Cloud-ready, scalable bioinformatics pipelines
• Clinical data science methodologies aligned with real-world healthcare applications

Together, these projects demonstrate an end-to-end approach to next-generation precision medicine—combining data-driven insight, biological interpretability, and clinical relevance.

Data Science & Predictive Analytics - Multi-Domain Project Portfolio

A collection of end-to-end analytics projects spanning real-world domains including marketing analytics, customer intelligence, cybersecurity, NLP, HR insights, fraud detection, energy analysis, ML education, compensation science, and startup analytics.

Each project is built using a consistent, reproducible workflow:

1. Data Cleaning & Feature Engineering
2. Exploratory Data Analysis (EDA)
3. Predictive Modeling (ML, statistical modeling, or NLP)
4. Insight Generation & Communication

The structure makes it easy to open any folder, understand the problem, run the notebook/script, and reproduce the full workflow. It’s both a practical learning resource and a demonstration of applied data science across diverse, high-impact domains.

Clinical Data Science & Health Analytics - Translational AI in Public Health Portfolio

A curated set of clinical data science and health analytics projects demonstrating real-world applications of machine learning, statistical modeling, and evidence-based analysis across healthcare. Projects span disease prediction and risk stratification, treatment adherence modeling, patient behavior analysis, and population-level health insights.

The portfolio includes:

• End-to-end predictive modeling pipelines for chronic disease, metabolic conditions, and clinical outcomes
• Risk scoring systems and stratification frameworks inspired by real-world clinical decision support tools
• Behavioral and lifestyle analytics that surface patient-level patterns influencing disease progression
• Treatment initiation and adherence modeling, plus data cleaning, feature engineering, and statistical validation for replicable, interpretable insights

Each project is organized for reuse and reproducibility, with clear structure, well-documented analysis steps, and modular code aligned with best practices in clinical AI and health informatics.

Infectious Disease Modelling & Epidemiology - Computational Transmission & Intervention Analytics Portfolio

A focused infectious disease modelling portfolio that turns transmission questions into reproducible code—pairing compartmental and stochastic perspectives with intervention-style analyses so dynamics can be explored before real-world deployment.

Key capabilities:

• Python, R, shell, and Jupyter-friendly workflows sized for transparent assumptions and iterative experimentation
• Transmission modelling motifs suited to vector-borne contexts such as malaria and related scenario exploration
• Modular repository layout so methods can extend to additional pathogens or geographical settings
• Open, version-controlled artifacts aligned with reproducibility expectations in computational epidemiology

The portfolio bridges mechanistic intuition with simulation discipline—stress-testing hypotheses with documented pipelines rather than one-off scripts.

Computational Biomedical Research - Integrated Imaging, Sequence & Oncology AI Portfolio

A multi-project computational biomedical laboratory spanning neurological signal analytics from eye-movement data, retinal-image cardiovascular risk stratification, DNA sequence–based gene expression modeling, oncology genomics and radiomics, ultrasound-informed hepatology analytics, microscopy cell phenotyping, and AI-guided drug and cancer-target discovery.

Key capabilities:

• Deep learning and classical ML stacks with diagnostics oriented toward interpretability and clinical relevance
• Imaging and sequence pipelines—from CNN-style fundus workflows to radiomics, ultrasound tasks, and sequence encoders
• Robust feature engineering, statistical validation, and artifact exports that keep experiments auditable
• Standalone project folders with notebooks and scripts that preserve reproducible structure end-to-end

Together these workflows show how modern ML can stay disciplined for biomedical rigor—from discovery through transparent evaluation and reporting.

Climate, Energy & Green Microbiology - Forecasting, Renewables & Sustainable Systems Portfolio

A climate and energy intelligence portfolio spanning AI-assisted global weather forecasting, renewable optimization under climate variability, electricity pricing strategies, load and demand forecasting, extreme-event risk modeling, reinforcement learning for energy systems control, and sustainable transition scenarios—including datasets and narratives aligned with green biotechnology contexts where climate intersects microbiology.

Key capabilities:

• Time-series and probabilistic forecasting layered with deep learning and graph-inspired atmospheric models
• Climate-informed renewable yield and planning analytics grounded in realistic variability assumptions
• Electricity market–aware modeling for price, dispatch, and volatility-aware decision support
• Scenario-centric tooling for decarbonization pathways and resilient infrastructure planning

The collection frames sustainability challenges as forecasting and operations problems—pairing disciplined data governance with models stakeholders can inspect and extend.

🌍 Why Smarter Algorithms Can Still Fail the Climate

A governance-focused essay on climate machine learning: why benchmark accuracy and innovation narratives are not enough when models sit inside electricity markets, supply chains, and institutions with misaligned incentives. Draws on high-impact ML-for-climate surveys and argues for auditable standards, independent evaluation, and outcome-based accountability—not just better predictions.

🛰️ Graph Neural Networks for Weather Forecasting: A Critical Analysis of GraphCast

A structured review of graph neural networks for medium-range global weather prediction, centered on the GraphCast paradigm: graph formulation, training on reanalysis data, skill versus NWP baselines, and open questions around uncertainty, extremes, and robustness under climate shift—with links to reproducible portfolio work on AI-based global forecasting.

🔮 A Bayesian Machine Learning Framework for Modeling Infectious Disease Outbreaks

Describes a Bayesian ML pipeline for epidemic time series: feature construction, probabilistic count models, posterior inference, and posterior predictive forecasts that expose uncertainty—contrasted with deterministic single-point predictions—plus discussion of strengths, limits, and links to an open implementation for outbreak prediction.

🏛️ AI Is Moving Faster Than Governments Can Respond

An AI governance essay distinguishing technical safety from institutional governance: anticipatory regulation, capacity gaps, voluntary “responsible AI” versus enforceable accountability, and concrete mechanisms—risk-tiered rules, audit infrastructure, and international coordination—grounded in leading policy and FAccT-era literature.

♻️ Can AI Fight Climate Change Without Worsening It?

Explores the tension between AI as climate infrastructure (grids, disasters, land use) and the hidden environmental cost of compute, hardware, and rebound effects; argues for lifecycle disclosure, impact assessment, equity in deployment, and governance that rewards verified emissions outcomes rather than green narratives alone.

👁️ Wearable Sensors for Brain Disorder Detection Through Eye Movements: A Comprehensive Research Review

How eye-tracking wearables are transforming neurological diagnostics from Parkinson's to Alzheimer's. This research review examines innovative wearable, high resolution eye movement sensors capable of detecting neurological disorders in real time, potentially reshaping how clinicians screen for conditions like Parkinson's disease, Alzheimer's disease, concussion, and traumatic brain injury.

🦟 Simulating and Fitting Malaria Transmission Model in Madagascar: Impact of Insecticide-Treated Nets

An applied research post on modeling disease transmission and evaluating ITN interventions using R. This study investigates the epidemiological impact of ITNs under varying resistance scenarios using a modified Susceptible, Infected, Recovered (SIR) model with integrated vector dynamics, simulating malaria transmission in Madagascar and comparing outcomes across different intervention scenarios.

🧬 Predicting Gene Expression from DNA Sequence Using Deep Learning Models

A breakthrough in computational biology exploring how advanced neural network architectures can learn the complex regulatory grammar embedded in DNA. This research demonstrates how artificial intelligence can decipher how genes are turned on or off a feat with profound implications for precision medicine, functional genomics, and therapeutic innovation.

📘 From Retina to Risk: Predicting Cardiovascular Health Through Deep Learning

A comprehensive analysis of Nature Biomedical Engineering (2018), exploring how retinal imaging and deep learning can predict systemic cardiovascular health. Highlights AI's role in early disease detection and precision health, demonstrating that deep learning models can predict key cardiovascular risk factors directly from retinal fundus photographs.