Skip to content

About This Research

Daniel Yarmoluk — AI Researcher & Systems Architect

I build AI systems that solve real operational problems. My background spans industrial IoT sensor networks, expert system design, and multi-agent orchestration across domains from medical device manufacturing to legal knowledge management.

This portfolio represents my approach to applied AI research: show working systems, measure real outcomes, and document reproducible methodologies that transfer across domains.

Research Philosophy: Show, Don't Tell

Traditional research portfolios present papers, abstracts, and theoretical frameworks. This portfolio takes a different approach — it IS the proof. Every chapter contains:

  • Working demonstrations you can interact with
  • Reproducible system architectures you can deploy
  • Measured outcomes tied to business value
  • Technical depth suitable for peer review

If you want to know whether I can build multi-agent AI systems for construction, the answer is in Brief 1. If you want to see computer vision for safety monitoring, Brief 3 has the architecture. The research speaks for itself.

What This Portfolio Demonstrates

1. Rapid Domain Mastery

On February 16, 2026, I systematized five unfamiliar technical domains in a single session — from oral surgery clinical protocols to peptide science for strength athletes. Each domain required:

  • Understanding complex technical literature
  • Building structured knowledge representations
  • Creating decision support tools
  • Validating accuracy against domain experts

This same capability applies to construction subdomains: mechanical systems design, structural load analysis, logistics optimization, or regulatory compliance. The methodology is domain-agnostic.

2. Multi-Agent AI Orchestration

Modern AI systems require orchestrating multiple specialized models — vision models for site photos, language models for specifications, graph models for BIM relationships. This portfolio demonstrates production-quality orchestration across:

  • Agent role definition and task decomposition
  • Inter-agent communication protocols
  • Quality control and validation layers
  • Human-in-the-loop intervention points

3. Knowledge Graph Architecture

Unstructured construction data (emails, PDFs, meeting notes) contains critical project intelligence that traditional databases miss. This research shows how to:

  • Extract entities and relationships from technical documents
  • Build domain-specific ontologies (IFC, COBie, Omniclass)
  • Enable semantic search and automated reasoning
  • Integrate with existing BIM and ERP systems

4. Technical Communication

AI research requires explaining complex systems to multiple audiences — engineers need architectural details, executives need ROI justification, project teams need practical workflows. Every brief in this portfolio balances technical rigor with operational clarity.

How to Use This Portfolio

Sequential Reading: Start with Brief 1 and progress through the research arc — from foundational architecture to deployment strategy to ROI measurement.

Topic-Based Navigation: Jump directly to briefs that align with your focus area. Each brief stands alone with its own context and references.

Interactive Exploration: Use the interactive tools embedded in each brief to test scenarios and explore edge cases.

Technical Deep Dives: Code examples, architecture diagrams, and system specifications are provided for implementation-level understanding.

Contact

Built With

This portfolio uses MkDocs Material and GitHub Pages — both open-source tools. The value is not in the publishing platform, but in the research methodology, system architecture, and demonstrated velocity.

The same frameworks shown here (multi-agent orchestration, knowledge graph construction, RAG pipelines) are production-ready for deployment in construction operations at scale.