Selected work

Professional users needed fast, accurate, citation-backed answers across complex documents.

A RAG architecture with domain-specific chunking, hybrid search, and grounded answer templates.

Reduced search time and created a controlled path toward a public product.

Decision-makers needed to understand long-term feedback effects of demographic policies.

CLD, stock-flow modeling, and scenario design for comparing policy interventions.

Converted qualitative concepts into discussable variables and testable policy structures.

Unplanned downtime and fragmented maintenance records increased operational cost.

Odoo-based CMMS implementation, process redesign, and data readiness for PdM.

Improved operational visibility and prepared the ground for predictive analytics.

The media space needed transparent criteria for source credibility and content health.

A rating model, information-flow analysis, and source-credibility scoring framework.

A product framework for media transparency and multi-layer fact-checking.

Evaluating different configurations of a multi-agent system before deployment required a controlled test environment.

Three-dimensional agent modeling with environmental physics and uncertainty inside a high-fidelity agent-based simulation engine.

Sensitivity maps and what-if scenarios to optimize layout and reduce decision risk.

Local grids lacked an economic and technical model for peer-to-peer energy trading; goal: reduce grid dependency and improve economic efficiency of distributed generation.

Three-layer architecture (producer / consumer / market-maker), real-time pricing via double-auction algorithm, and smart contracts for automated energy transaction settlement.

Operational pilot at 12-building scale; scalable architecture ready for district or industrial-zone deployment.

Lack of trust and transparency among supply-chain actors created cost and error.

Immutable recording of logistics transactions and smart contracts to remove intermediaries.

End-to-end shipment traceability and less human error and opportunity for fraud.

Investment decision-makers needed early detection of emerging technologies.

Semantic analysis of papers and patents with NLP and LLMs to detect growth patterns and field actors.

Trend forecasts and actor-network visualizations for investment-risk management.