Module 1: Foundations of Data Analysis and Machine Learning
Biomedical data foundations: data types, study types, descriptive statistics and visualization.
Data preparation: working with pandas, data cleaning, missing values.
Applied statistics: group comparisons, correlations, linear and logistic regression, correlation vs. prediction.
Intro to machine learning: supervised vs. unsupervised learning, features and labels, train/test split.
Prediction and classification models (scikit-learn): linear and logistic regression, decision trees, Random Forest, KNN.
Model evaluation: model comparison, sensitivity, specificity, PPV/NPV, ROC and AUC, overfitting, class imbalance, probability calibration.
Methodological pitfalls: confounding, selection bias, data leakage and batch effects.
Capstone exercise: analyze a biological dataset, build a model, evaluate performance and present the limitations of the analysis.
Module 2: Cloud and Engineering Practices on Google Cloud Platform
GCP fundamentals: projects, permissions (IAM), billing and cost control.
Google Cloud Storage: buckets, uploading and downloading files, organizing data.
BigQuery: SQL queries, working with public datasets, loading tables.
Cloud-connected Colab: reading from GCS and BigQuery in a notebook and running analyses on cloud data.
Git and reproducibility: version control, GitHub, organizing notebooks and code for sharing and reproducibility.
Privacy and access control: anonymization, permissions and information security when working with sensitive data.
Module 3: Biomedical Data and Studies
Study types: cohort studies, retrospective vs. prospective, clinical trials and real-world studies.
Health data: EHR, claims, medical coding systems and interoperability (FHIR).
Omics data: genomics, transcriptomics, single-cell and multi-omics, and the challenges of working with them: missing data, batch effects, high dimensionality.
Clinical trials and real-world evidence: RWD and RWE, trial outputs.
Lab, longitudinal and patient-generated data: longitudinal data, patient-reported outcomes (PROs), sensors and wearables.
Imaging and clinical text (overview): data types and processing challenges.
Data quality and bias: noisy labels and dataset shift.
Hands-on SQL and analysis in BigQuery.
Module 4: Generative AI and Large Language Models
Neural networks, Transformers and embeddings: introduction to deep learning, tokens and context windows, vector representations, foundation models.
Systematic prompting and structured output: role and context, examples, constraints, structured output, iterative improvement.
Tool calling, RAG and vector databases: function calling, information retrieval, building a RAG system over scientific literature.
AI agents: principles and examples.
Multimodal models: introduction and use cases.
System evaluation and hallucinations: model hallucinations, evaluation methods, external validation.
Responsible AI and regulation: ethical and legal issues, human oversight.
Tools: Gemini, Google AI Studio, Vertex AI, NotebookLM, ChatGPT and API-based prototyping.
Module 5: Building AI Products
User discovery and workflow mapping.
MVP and product requirements: defining a minimum viable product and writing product requirements.
Build vs. buy: deciding between developing in-house and adopting an existing solution.
Human-in-the-Loop and usability: designing systems with human oversight and intervention.
KPIs, A/B tests and value: key performance indicators, controlled experiments, articulating business value.
PRD workshop: each participant writes a basic AI product requirements document.
Module 6: AI in Industry and Career Integration
Drug discovery.
Genomics and Omics: AI applications in Omics data.
Digital pathology and medical imaging: AI applications in imaging.
Clinical trials: optimization and patient recruitment.
Precision medicine.
Digital biomarkers and clinical decision support.
Career panel: meet industry professionals: roles, career-transition paths and what companies look for.
Module 7: Capstone Project and Portfolio
Teams work on an applied challenge using a relevant dataset. Each team delivers:
A working prototype or live notebook in a clean GitHub repository.
An evaluation of the solution’s performance.
A product and business rationale, including a one-page summary for a non-scientific audience.
A five-minute professional pitch to industry representatives.
The project becomes your portfolio piece.
*Subject to changes.