I am a computational chemist and AI researcher at the University of Glasgow working at the intersection of machine learning, large language models, and chemical discovery.
My work focuses on building autonomous, agent-based systems that integrate retrieval, reasoning, and tool-use to solve complex scientific problems. I design and deploy AI systems that connect:
- Unstructured knowledge (scientific literature)
- Structured workflows (molecular design, synthesis, experimentation)
I am particularly interested in:
- LLM-driven reasoning and multi-agent systems
- Generative models for molecular design
- End-to-end AI pipelines for real-world discovery
- Bridging AI systems with experimental and robotic platforms
-
ACRA — Autonomous Chemical Reasoning Agent
Multi-agent LLM system translating scientific literature into executable synthesis workflows
→ RAG + planning + tool-use + self-correction -
MIDAS
Language-controlled 3D molecular generation with spatial reasoning
→ LLMs interacting directly with molecular representations -
LangSim
Agent-based framework for automated materials discovery
→ Orchestrates multi-step scientific workflows
- LLMs and agent-based systems for scientific reasoning
- Machine learning for molecular generation and property prediction
- Chemical space exploration and drug design (LBDD / SBDD)
- Autonomous and closed-loop discovery systems
- Scientific software, data pipelines, and reproducibility
Programming Languages
Machine Learning & AI Systems
- Multi-agent LLM systems (ReAct, tool-use, planning)
- Retrieval-Augmented Generation (RAG), vector databases
- LLM evaluation and self-correction (LLM-as-a-judge)
Computational Chemistry
- Ligand- and structure-based drug design
- Molecular dynamics and simulation
- Chemical dataset curation and analysis
|
Midas Agentic interface for language-controlled 3D molecular generation and manipulation 
|
|
ACRA Multi-agent LLM workflow for automated synthesis planning and execution from scientific literature 
|
|
Spalt Alignment of Ligand Topographies using molecular surface features. 
|
|
Molecular Spaces Quantification of selection in chemical spaces and denovo molecular generation using Assembly Theory. 
|
|
Chemputation Chemputer and Chemputation -- A Universal Chemical Compound Synthesis Machine. 
|
|
molgrep CLI SMARTS based molecule retrieval 
|
|
MolNCA Neural Cellular Automata for molecule generation. 
|
|
LangSim LLM-driven agent framework for orchestrating materials simulation workflows 
|
|
AgentLabs Framework for autonomous lab agents and experimentation.
|
- GitHub: https://github.com/pagel-s
- Google Scholar: https://scholar.google.com/citations?hl=en&user=cq0VR8wAAAAJ
- LinkedIn: https://www.linkedin.com/in/sebastian-pagel-5a0b71148/