Foresight Logic

Intelligent Software & Services

Providing bespoke analytics, predictive models, and intelligent solutions
for healthcare, finance, industry, and research since 1998.

Rooted in healthcare research, psychometrics, and decision support — fields where the analysis has to be right, the methods have to be transparent, and the systems have to keep running long after the kickoff meeting.


Measure Carefully – Model Honestly – Build Durably

Skills

  • LLM fine-tuning and workflow integration. Intelligent systems design. Agentic workflow development and deployment.

    Leveraging emerging advances as well as the tried and true.

  • Experiments, surveys, audits, evaluation, cross-sectional and longitudinal analysis.

    From study design through final reporting.

  • Predictive, causal, and psychometric models built to hold up under scrutiny — and keep working in production.

    Validation, calibration, and documentation included.

  • Purpose-driven model selection, harnessing. workflow design, and evaluation for AI coding agents and non-coding agentic systems.

    Built so the agents work the way you intended.

Methods

  • Start from the decision.
    Before touching the data, know what choices the solution is supposed to inform.

  • Match method to problem.
    A logistic regression you can explain often beats an ensemble you can't. Escalate complexity only when it earns its keep.

  • Build for the second year.
    Most of the cost of analytics shows up after launch. Design and document with that horizon in mind.

  • Know what you don't know.
    If the data won't support the question, say so early. Good decisions cannot be made on analysis that hides its weaknesses.

  • Be tool-agnostic.
    No vendor pitches, no favorite framework. Python, R, SQL, Stata, C#, even Excel — Pick the tool that fits the problem and the team that has to live with it.

  • Build to defend.
    Documentation, tests, and a clear chain of reasoning ship with every deliverable. If your model lands in front of a regulator or a board, you can explain it.

  • Senior, by default.
    You work directly with the people doing the work. No layered staffing, no handoffs to juniors after the kickoff.

Results

From select engagements…

  • VP AI/ML & Advanced Analytics (2022 - 2026)

    Designed and developed patent pending AI/ML technology and workflow for automating the medical coding of clinical notes. Developed comprehensive analytics workflows to support operations and R&D.

  • (2005 - Present)

    Methodologist, statistician, and analytic programmer on studies and decision-support tools covering ADHD, autism-spectrum, developmental disability and depression. Co-author on multiple papers.

  • (2016 - Present)
    Statistical consultant on diagnostic screening of adolescent psychopathology. Multi-year collaboration producing peer-reviewed research on the Pediatric Symptom Checklist, depression and suicidal-ideation screening, and longitudinal outcomes. Co-author on multiple papers.

  • Evaluator on the CMS Health Care Innovation Award "OneCare" project. Psychometric consultant on behavioral-health screening.

  • Biostatistician and analytic programmer on the molecular basis of Spinocerebellar Ataxia Type 1 using RNA-Seq data; co-author on resulting research in Neuron.

  • Director, AI/ML & Advanced Analytics (2020-Present)

    Consult to business and technical leadership on emerging trends and best practices in the application of AI to optimize solutions to customer challenges.

    Design & develop bespoke AI/ML and analytic solutions for organizations primarily in the healthcare ecosystem.

  • Statistical consultant on development of ambulatory glucose profiles and analysis of diurnal glucose oscillations.

  • Designed and applied a fuzzy multi-branching tree algorithm for customer cross-sell solicitation targeting. Response rates ran more than three times the industry average.

  • Built a fuzzy case-based reasoning engine to predict whether 90-day-delinquent homeowners would foreclose or cure. Significantly outperformed in-house statistical models.

  • Applied machine-learning algorithms to evaluate books of business in the home-mortgage industry.

Founder

Foresight Logic Is Paul Bergmann

Foresight Logic is the consulting practice of Paul Bergmann — AI/ML solutions architect, statistician, and psychometrician.

Paul founded Foresight Logic in 1998 to bring serious statistical and machine-learning capabilities to organizations that needed defensible analysis without the overhead of a large consultancy.

Paul provides a customized, tailor-made approach to his work with clients — including academic medical centers, healthcare systems, financial institutions, and research groups — many of whom he has maintained relationships with for years through an approach that combines technical expertise with clarity and candor.

His technical roots include training in educational psychology and mathematics at the University of Minnesota. He moved early into the application of fuzzy logic, case-based reasoning, recursive partitioning, and neural networks at a time when most of the field was still calling them "expert systems" and “data mining.” That work began in clinical research and behavioral health in the early 1990s and expanded into engagements with clients like Anderson Consulting's Advanced Technologies Group (now Accenture), Mercantile Bank, Residential Funding Corporation, and Credit Suisse First Boston.

His work over the past number of years has focused on helping clients navigate the constant advancement of Artificial Intelligence models, tools, and infrastructure to establish purpose-driven best practices that are accessible and sustainable.

“I started Foresight Logic to do serious analytic work and advance the application of emerging technologies to real world challenges.

Twenty-five years in, I’m convinced it was the right call. The clients who come back are the ones who want the senior person on the problem from the first meeting to the last line of code — and that's the only kind of engagement I run.

If that's the kind of partner you're looking for, send a note. I read every message.”

– Paul Bergmann

Published Work

A practice rooted in research stays sharp on method. Foresight Logic's work appears in pediatric, psychiatric, and methodological journals. Here is a selection of high-impact pieces:

  • Murphy, J., Bergmann, P., Chiang, C., Sturner, R., Howard, B., Abel, M., Jellinek, M. (2016). The PSC-17: Subscale Scores, Reliability, and Factor Structure in a New National Sample. Pediatrics, 138(3).

  • Sturner, R., Howard, B., Bergmann, P., et al. (2016). Autism Screening with Online Decision Support by Primary Care Pediatricians Aided by M-CHAT/F. Pediatrics, 138(3).

  • Ingram, M., et al., Bergmann, P., et al., Orr, H. T. (2016). Cerebellar Transcriptome Profiles of ATXN1 Transgenic Mice Reveal SCA1 Disease Progression and Protection Pathways. Neuron, 89(6), 1194–1207.

  • Jellinek, M., Bergmann, P., Holcomb, J. M., et al. (2021). Recognizing Adolescent Depression with Parent- and Youth-Report Screens in Pediatric Primary Care. Journal of Pediatrics, 233.

  • Sturner, R., Bergmann, P., Howard, B., et al. (2022). Do Autism-Specific and General Developmental Screens Have Complementary Clinical Value? Journal of Autism and Developmental Disorders.

  • Bergmann, P., Lucke, C., Nguyen, T., Jellinek, M., Murphy, J. M. (2018). Identification and Utility of a Short Form of the Pediatric Symptom Checklist Youth Self-Report. European Journal of Psychological Assessment.

  • Patent Pending: Automatic Medical Coding Determination. M&G 19491.0001US01. U.S. Application No. 18/765,093

Grounding Truths

Pragmatism is a discipline.
The biggest model is rarely the right one. I start from the decision you're trying to make and work backward to the simplest, most defensible solution that gets you there.

Explainability isn't optional.
A model nobody understands is a liability waiting to happen. My work ships with the documentation, diagnostics, and plain-language reasoning that let you stand behind it.

Solutions should outlive the ones who build them.
Planned obsolescence should only apply to consultants, not products. I will work with you to ensure a solution that is maintainable and sustainable.

Right tool, not favorite tool.
I have no quota to fill on any platform, language, or vendor. The only question I ask is what your team will be able to maintain after I leave.

A Mixture of Experts

Together, we cover the full depth and breadth of modern tech stacks and intelligent solutions.

How Can I Help?

Send a note describing what you’re looking to do, and I will get back to you promptly.