Intelligent planning / Agentic AI and optimization
Agentic AI and Mathematical Optimization
Bringing language-model reasoning and constrained optimization together for industrial planning.
Problem statement
Connect domain knowledge with planning decisions that must satisfy operational constraints.
My contribution
Cross-functional development of an agentic planning approach that connects retrieval, LLM reasoning, and mixed-integer optimization.
Technical challenges
Domain-level considerations, without implementation-specific details.
- Grounding reasoning in relevant domain context.
- Distinguishing plausible suggestions from feasible plans.
- Keeping planning objectives and constraints explicit.
Methodology
Pair retrieval-augmented generation and LLM-based reasoning with mixed-integer programming.
- Retrieval-augmented generation to provide domain context for reasoning.
- Mixed-integer programming to express planning objectives and constraints.
Conceptual architecture
Planning question
Context and reasoning
Constrained optimization
Candidate plan and review
Technologies and methods
- Agentic AI
- RAG
- Optimization