This site may give you a flavor of the the kind of problems I am interested in, but there is much to add and this is a work in progress. I spent a long time developing software for business tasks in various domains, and I have a strong interest in the intersection of data science and software engineering. I can work with your data science team on the following tasks:

• Leveraging data science to extract actionable insights from operational data. This encompasses a wide range of activities, from exploratory data analysis to developing machine learning models using data from operational systems or analytics platforms. Typical challenges I can help address include:

  • Predicting customer behaviors such as churn, demand, or product returns
  • Monitoring key behaviors, actions, or performance metrics
  • Building descriptive models for customer or process segmentation
  • Creating prescriptive models that recommend actions to prevent negative outcomes or improve the likelihood of success—for example, optimizing help desk ticket triage

• For practical examples, see this repository. Effectively framing a business challenge as a data science problem requires experience, careful analysis, and close collaboration. Automated or “cookbook” approaches often fall short, leading to models that are difficult to interpret or deploy in production. I can partner with your team to guide the initial iterations of your machine learning or optimization initiatives—helping you identify the most relevant data sources, select suitable modeling techniques, and ensure that results are both interpretable and actionable. This approach increases the likelihood of building solutions that deliver real business value and are ready for operational use.

• Managing model complexity and evolution. Determining the appropriate level of complexity for a model is crucial to solving business problems effectively—overly simple models may miss important patterns, while overly complex ones can be difficult to interpret and maintain. I can help assess and refine models over time, balancing accuracy, interpretability, and maintainability. Adopting a satisficing approach ensures that models are “good enough” for the task at hand, without unnecessary complication. For an example of this work, see this publication.

• Summarizing and profiling your data assets to guide problem-solving. Organizations accumulate large volumes of data, but extracting meaningful insights requires systematic exploration and summarization. This ongoing process ensures that valuable information is incorporated into each iteration of your models. Recent advances in data profiling and summarization techniques (see this playlist) offer new ways to surface actionable insights. I can help you select and apply the most suitable methods for your specific data assets and business objectives.

• Collaborating with your optimization team to integrate machine learning and statistical models with optimization workflows. Many real-world planning and decision problems involve uncertainty—such as unpredictable demand, weather, or supply disruptions. Machine learning models can quantify this uncertainty by generating forecasts or probability distributions, which can then be incorporated into optimization models to improve planning and resource allocation. For example, agricultural businesses can use weather prediction models to inform crop planning, while retailers can leverage demand forecasts to optimize inventory for perishable goods. Advances in optimization tools have made it increasingly practical to embed data-driven predictions into stochastic optimization frameworks, enabling more robust and adaptive decision-making.