Building intelligent systems at the intersection of machine learning, topology, and LLMs

I’m an AI Researcher currently working on LLM-based multi-agent systems for deep research and legal AI.

• 🧠 Designing multi-agent LLM pipelines (LangGraph, LangChain, OpenAI, Azure)
• 🔬 Researching topological data analysis (persistent homology) in ML
• 📰 Scientific Papers
  • Topology-Informed Machine Learning for Efficient Prediction of Solid Oxide Fuel Cell Electrode Polarization
  • Persistence is All You Need - A Topological Lens on Microstructural Characterization
  • Computational Topology in Energy Engineering: A New Lens on Microstructural Complexity
• 🏆 Delivered state-of-the-art results on graph benchmarks (MUTAG, PROTEINS, NCI1)
• ⚡ Built systems with 10× efficiency improvements in scientific simulations

• Machine Learning & Deep Learning
• LLM Systems & Multi-Agent Architectures
• Topological Data Analysis (TDA)
• Graph Neural Networks
• Scientific Computing & Optimization

• Designed a deep research system combining retrieval, reasoning, and metadata-aware agents
• Built with LangGraph + OpenAI + Elasticsearch
• Improved evaluation metrics on internal benchmarks
• Developed Streamlit interfaces + evaluation pipelines

• First-author research on persistent homology in deep learning
• Predicted microstructural properties with high accuracy:
  • R² up to 0.91
  • Pearson r up to 0.96
• Reduced simulation time from 1 hour → 1 minute

• Introduced 2-Lipschitz stable persistence transformation
• Integrated topology into contrastive GNN pipelines
• Achieved SOTA:
  • MUTAG: 90.43%
  • PROTEINS: 86.14%
  • NCI1: 85.93%

• 💼 LinkedIn: https://linkedin.com/in/maksym-szemer
• 🧬 ORCID: https://orcid.org/0009-0005-0510-7610