RavenLLM

University of Illinois at Urbana-Champaign

Computational Chemistry Lab

Examples: "negative photoresist with 405nm sensitivity", "two-photon polymerization resin", "PMMA-based e-beam resist", "high-contrast i-line photoresist"

Ready to design photoresist materials

Specify target properties or polymer structures for two-photon lithography

Model Architecture

Base Model: Qwen-2.5-7B-Instruct
Fine-tuning: Unsloth (4-bit LoRA)
Training Data: 200k chemical library
Deployment: Ollama + HuggingFace
Inference: Quantized (GGUF)

Performance Metrics

Chemical Accuracy 94.2%
Synthesis Feasibility 89.7%
Fine-tuning Loss 0.23

Benchmarked on photoresist synthesis & polymer design tasks

Citation

Li, R., Ceballos, D., Funke, O., Ogunjobi, O., Eshan, M., Zeng, Y. "RavenLLM: Transformer-Based Retrosynthetic Planning and Photoresist Discovery for Autonomous Materials Design." IBM-Illinois Discovery Accelerator Institute, ECE Department, UIUC (2025)

Research Team

Dr. Renjie Li, PhD

Principal Investigator

AI for Autonomous Materials Discovery

Undergraduate Researchers:

Daniel Ceballos

AI/LLM for Polymer Design

Owen Funke

Neural Networks & Chemical Synthesis

Olamide Ogunjobi

LLM Integration & Photoresist Design

Mohamed Eshan

AI-Driven Photosensitive Materials

Yinqi Zeng

Electrical Engineering & ML Systems

IBM-Illinois Discovery Accelerator Institute
ECE Department, UIUC