AI/SI Portfolio

Human-AI/SI Collaboration • AI/SI Assurance • Independent Researcher

I conduct AI/SI-augmented research and apply AI/SI assurance methods to AI/SI-assisted workflows, using ChatGPT and other large language models (LLMs) as iterative analytical collaborators. My methodology integrates human-led research questions, source investigation and analysis, AI/SI-augmented analysis, human evaluation and redirection, source and evidence validation, risk classification, root cause analysis, mitigation verification, synthesis, interpretation, and the development of evidence-based findings and recommendations.

I apply this methodology to complex research, information-analysis, assurance, and decision-support problems. It builds on a multidisciplinary foundation in independent research, scientific analysis, enterprise technology, root cause analysis, information quality, and data analytics.

AI-Augmented Research
Human-led research questions, source investigation & analysis → AI/SI-augmented analysis → Human evaluation & redirection → Human source & evidence validation → Human–AI/SI synthesis → Human interpretation → Evidence-based findings & recommendations

My approach uses AI/SI to extend human analytical capacity rather than replace human expertise. Through iterative Human-in-the-Loop (HITL) ↔ AI/SI-in-the-Loop (AIITL) collaboration, I apply human judgment throughout the research process, from research questions, source investigation, and analysis through evaluation and redirection, source and evidence validation, synthesis, interpretation, and the development of findings and recommendations. 

I have applied this approach to real-world projects, including AI-augmented collection research and acquisition recommendations spanning the complete nonfiction catalog, evidence-based collection-weeding guidance, and emerging subject areas for long-term planning of a multi-branch public library system.

My work in generative AI/SI began in 2022 and includes AI/SI-augmented and AI-assisted research, AI/SI model training and evaluation, prompt engineering, custom GPT development, structured AI/SI workflows, human validation, information quality, and agentic AI/SI experimentation.The work reflects an approach in which human judgment operates throughout the entire research and analytical process while generative AI/SI extends analytical capacity.

I am also extending this methodology into AI/SI assurance: the systematic collection and evaluation of evidence concerning AI/SI behavior, risk classification, root cause analysis, corrective-action tracking, and verification that mitigations effectively address identified failures. This work emphasizes the distinction between implementing a corrective action and demonstrating through evaluation that the action actually reduces risk.

These capabilities build on an extensive enterprise technology and independent research foundation spanning information analysis, SQL development and analytics, enterprise data, business intelligence, data and information quality, Python scientific computing, systems and database administration, security administration, and decision-support analytics.

Featured Areas

The portfolio brings together five complementary areas of research, AI/SI, information analysis, and technology. My broader independent research experience includes literature and library research, observational analysis, scientific datasets, evidence synthesis, and computational analysis.

AI/SI-Augmented Research & Decision Support

Human-led applied research and iterative Human-in-the-Loop (HITL) ↔ AI/SI-in-the-Loop (AIITL) collaboration to investigate complex questions, evaluate information and evidence, refine analysis, synthesize findings, validate results, and evidence-based decisions. 

AI/SI Assurance, Risk & Root Cause Analysis

Evidence collection and validation, AI/SI failure analysis, risk classification, root cause and contributing-factor analysis, corrective-action tracking, mitigation verification, residual-risk assessment, and the development of defensible assurance findings.

AI/SI Model Training, Evaluation & Human Validation

AI/SI model training and evaluation, Human-in-the-Loop (HITL) scientific analysis and validation, AI/SI response evaluation, factual and evidence validation, information quality, scientific classification and categorization, pattern recognition, and iterative output refinement.

Applied Generative AI/SI

Prompt engineering, custom GPT development, structured AI/SI workflows, large language model (LLM) applications across multiple platforms, and experimentation with agentic AI/SI for research, analytical, and knowledge-work applications.

Data, Information & Analytics

Enterprise SQL development, reporting, Business Intelligence (BI), analytics, data and information quality, metadata validation, information management, analytical problem-solving, and decision-support systems.

Independent Research & Publications

Independent interdisciplinary research incorporating literature and library research, observational analysis, scientific datasets, Python scientific computing, information analysis, evidence synthesis, and structured research methodologies. My eight-paper CERN Zenodo research corpus documents investigations into structured light phenomena and resonant field dynamics reflects a broader foundation in independent research, scientific classification, data analysis, and evidence-based inquiry.

Astronomical Observation, Scientific Classification & Earlier HITL

My experience with Human-in-the-Loop (HITL) scientific research and analysis predates the current generative AI/SI era. For more than a decade, I applied astronomical knowledge, observational judgment, pattern recognition, classification, and human validation to complex scientific imagery and datasets through NASA-funded Planetary Science Institute academic research projects, including Cosmoquest and CosmoquestX Data Mappers projects.

This work included independently applying observational judgment and scientific classification criteria to identify, classify, and validate phenomena missed by automated and machine-learning systems, supplementing machine analysis with human expertise. Human findings were also used as training data to improve subsequent machine-learning detection and classification.

Long-term telescopic astronomical observation and the analysis of astronomical, planetary, solar, satellite, spacecraft, and space-telescope imagery further developed the pattern recognition, classification, evidence evaluation, and analytical judgment that continue to inform my current research and AI/SI work.

Explore Astronomical Observation, Classification & Research Projects

Applied Research Example: Library Collection Development

A recent project for a multi-branch public library system demonstrates this methodology in practice. At the request of the Director of Library Systems, I conducted AI-augmented research across the complete nonfiction catalog to develop contemporary acquisition recommendations for library leadership. 

Using iterative Human-in-the-Loop (HITL) ↔ AI/SI-in-the-Loop (AIITL) collaboration and a structured, evidence-based research methodology, I investigated and evaluated potential acquisitions, directed and refined AI/SI-assisted analysis, performed source, bibliographic, metadata, and information-quality validation, and synthesized and interpreted the validated research into an 11-page decision-support report containing acquisition recommendations, collection-weeding guidance, and emerging subject areas for long-term planning.

Applied Research Example: Independent Scientific Research

In 2025, my research into structured light phenomena and resonant field dynamics culminated in a eight-paper research corpus published through CERN's Zenodo repository. The methodologies developed through that work continue to inform my approach to investigating complex questions, evaluating evidence, testing interpretations, recognizing patterns and inconsistencies, and distinguishing observation and established evidence from inference.

Across enterprise technology, astronomical observation and scientific classification, independent research, and generative AI/SI, one principle has remained constant: applying human analytical judgment to complex information to develop accurate, reliable, and evidence-based knowledge that supports understanding and informed decision-making.

Retrospective Assurance Example: Assuring AI/SI-Augmented Nonfiction Collection Recommendations

AI/SI-assisted research covering approximately 80 titles within each Dewey Decimal class and producing 50 final recommendations across the complete nonfiction catalog.

• Principal risks

  • • Incorrect titles, authors, dates, editions, or publication status.

    • Duplicate recommendations across categories.

    • Unsupported claims about authority or relevance.

    • Overrepresentation or omission of subject areas.

    • Recommendations inconsistent with collection objectives.

    • Plausible AI-generated explanations unsupported by evidence.

• Evidence collected

  • • Publisher and authoritative bibliographic records.

    • Existing catalog holdings.

    • Candidate and final recommendation lists.

    • AI-generated rationales.

    • Correction and consolidation decisions.

    • Collection-development criteria.

• Root cause categories

  • • Ambiguous or incomplete source metadata.

    • AI/SI conflation of editions or similarly titled works.

    • Separate analyses producing duplicate candidates.

    • Insufficiently constrained recommendation criteria.

    • Unsupported inference presented as fact.

    • Inadequate cross-category comparison.

• Corrective controls

  • • Independent bibliographic verification.

    • Cross-category duplicate detection.

    • Explicit evidence and evaluation criteria.

    • Human review of authority, accuracy, accessibility, relevance, diversity, and community value.

    • Separation of verified facts from analytical recommendations.

    • Final consolidated review before delivery.

• Mitigation verification

  • • Recheck every final title against authoritative records.

    • Confirm that each recommendation is unique.

    • Trace each recommendation to its supporting evidence.

    • Test whether the same controls catch deliberately introduced metadata and duplication errors.

    • Document remaining uncertainty rather than treating absence of detected errors as proof of perfection.

• Demonstrated outcome
  A validated, consolidated set of 80 contemporary nonfiction acquisition recommendations, accompanied by evidence-based weeding guidance and an emerging trends for library leadership decision support.

Interactive AI/SI Applications

OpenAI Custom GPTs

Developed and deployed a collection of domain-specific GPT applications that combine prompt engineering, structured reasoning, and AI/SI-assisted workflows to support research, education, and creative exploration. Each application demonstrates a different approach to applying large language models to specialized knowledge domains.

• • "Cosmic Quest Adventure": An interactive research simulation that combines deductive reasoning, scientific exploration, and problem-solving through virtual missions inspired by astronomy and astrophysics.

  • "Cosmic Weaver": A visual ideation assistant designed to explore field-based interactions, spatial relationships, and conceptual modeling through AI/SI-assisted creative workflows and original artwork.

  • "Astro Light Explorer": An expert-level astronomy assistant that supports the interpretation of photometric observations, atmospheric optical phenomena, and structured light patterns through AI/SI-assisted analysis and scientific reasoning.

Custom OpenAI GPT Screenshots

Autonomous AI/SI Workflows

xAI Grok Agentic Automations

Unlike interactive AI/SI applications, these autonomous workflows execute on recurring schedules, synthesize domain-specific information, and automatically generate structured reports for ongoing research and analysis.

Designed and deployed autonomous xAI Grok workflows for scientific reporting and publication analytics.

Project 1: Daily Astronomical Sun & Moon Times

Overview

This automation workflow executes on a daily schedule, retrieves astronomical ephemeris information for a specified observing location, synthesizes the results into a human-readable observational summary, and delivers the report automatically by email. The underlying prompt architecture is proprietary.

Project Summary

• Category: Agentic AI/SI Workflow

• Platform: xAI Grok

• Status: Deployed

• Delivery: Scheduled Email Automation

• Primary Skills: Prompt Engineering • Workflow Automation • Knowledge Synthesis

Project Type

• Agentic AI/SI Automation

• Scheduled AI/SI Workflow

• Automated Scientific Information Delivery

Objective

Develop an autonomous AI/SI workflow that generates and emails a daily astronomical observing report containing solar and lunar ephemeris information for a specified observing location.

Capabilities

• Executes automatically on a daily schedule.

• Retrieves current astronomical ephemeris data.

• Generates a structured natural-language summary.

• Calculates sunrise and sunset times.

• Reports moonrise, moonset, and lunar illumination.

• Provides observational context for planning astronomy sessions.

• Delivers the report automatically by email.

Skills Demonstrated

• Agentic AI/SI workflow design

• Prompt engineering

• Workflow automation

• Structured data interpretation

• Technical communication

• Astronomy domain knowledge

• Automated report generation

Representative Results

The following representative screenshots demonstrate the deployed Astronomical Sun & Moon Times workflow and illustrate the automated generation and email delivery of structured astronomical observing reports. Proprietary prompt engineering and implementation details have been intentionally omitted. The automation screenshots provide:

• Sunrise and sunset

• Moonrise and moonset

• Lunar illumination percentage

• Daylight duration

• Observing conditions

• Narrative interpretation suitable for planning observations

Implementation Notes

The underlying prompt architecture, workflow logic, and optimization techniques are proprietary. Screenshots are provided to demonstrate functionality and output quality while protecting implementation details.

The following representative screenshots demonstrate engineering and automation of the of the Astronomical Sun & Moon Times project workflow.

Agentic AI/SI Workflows

xAI Grok Automations

Project 2: Daily Zenodo Publications Analytics 

Overview

This autonomous AI/SI workflow monitors my published Zenodo research corpus on a recurring schedule, analyzes publication metadata, generates concise summaries of individual works, and compiles the results into a structured research report delivered automatically by email. The workflow streamlines ongoing publication monitoring while preserving proprietary prompt engineering and implementation details.

Platform

• xAI Grok

Project Type

• Agentic AI/SI Automation

• Research Analytics

• Scholarly Publication Intelligence

Objective

Develop an autonomous AI/SI workflow that monitors and summarizes a research publication corpus, transforming technical academic metadata into concise, readable research analytics delivered automatically.

Zenodo Publications Analytics Project Workflow

Autonomous workflow that analyzes a scholarly publication corpus and generates structured publication summaries for ongoing research monitoring.

Capabilities

• Executes on a recurring schedule.

• Reviews the current Zenodo corpus.

• Identifies publications and associated metadata.

• Generates concise summaries of each publication.

• Produces a structured research overview.

• Delivers the report automatically via email.

Technical Highlights

• Agentic AI/SI workflow design

• Prompt engineering

• Scientific literature summarization

• Metadata interpretation

• Automated research analytics

• Structured report generation

• Technical communication

• Autonomous scheduled execution

Skills Demonstrated

• Agentic AI/SI workflow design

• Prompt engineering

• Workflow automation

• Scientific literature summarization

• Research metadata analysis

• Knowledge synthesis

• Technical writing

• Automated analytics and reporting

Representative Results

The workflow produces a concise summary of published research, including publication titles, DOI references, and technical descriptions suitable for rapid review and portfolio monitoring.

Implementation Note

The underlying prompt architecture, workflow logic, and optimization techniques are proprietary. Screenshots are provided to demonstrate functionality and output quality while protecting implementation details.

The following representative screenshots demonstrate engineering and automation of the of Zenodo Publications Analytics project workflow.

Technologies & Platforms

OpenAI Custom GPTs

xAI Grok

Prompt Engineering

Agentic AI/SI

Workflow Automation

Research Analytics

Scientific Data Interpretation

Knowledge Synthesis

Technical Documentation

AI/SI-Assisted Research