Computational scientist building ML and simulation tools for molecular and materials problems.
PhD, Chemical Engineering (University at Buffalo) · Postdoc at NC State University Applied ML Scientist — molecules, materials, and proteins
I work at the intersection of machine learning, cheminformatics, and molecular simulation — building computational workflows that help experimental scientists prioritize what to test next. Whether that's screening 20M peptide candidates down to a tractable set, or using MD and free-energy methods to rank molecular interactions, my specialty is rigorous evaluation — benchmarking whether ML methods actually work before they ship, not just whether they run.
- 🔬 High-throughput peptide screening pipelines combining unsupervised ML with molecular dynamics
- 🤖 Agentic AI for computational chemistry — LLM tool-harness workflows (MCP / Claude Skills) that automate protein structure preparation and simulation setup — see CHARMM-GUI PDB Reader skill
- 🛠️ Open-source scientific software — core developer and maintainer of ChemML, and co-developer of FK-eABF, a PLUMED enhanced-sampling plugin that recovers converged free-energy landscapes from the earliest stages of sampling
- 📄 Published protein & biomaterials MD in 2026: silaffin biomineralization (JPC B), a force-kernel enhanced-sampling method (JCTC), and ELP phase behavior (Soft Matter)
- Model evaluation, uncertainty quantification, and XAI for chemistry (DeepSHAP/LIME/LRP) — my core specialty; see my ML StarterKit
- Molecular property prediction, transfer learning, and GNNs
- Protein language model embeddings (ESM-2, ProtBert) for peptide design
- Molecular dynamics, enhanced sampling, and free-energy methods (GROMACS/PLUMED)
- Cheminformatics (RDKit) and molecular representations
- Scalable unsupervised learning for large sequence and molecular datasets
- Sonpal, Pradhan, Hachmann — Equally Valid, Differently Readable: A Cross-Domain Benchmark of Post-hoc XAI for Molecular and Materials Property Prediction (in preparation)
- Sonpal, Pfaendtner — Phosphoserine Charge State Drives Ion Condensation and Spatial Polyamine Presentation in Multi-Repeat Silaffin. The Journal of Physical Chemistry B (2026)
- Sonpal, Afzal, An, Chandrasekaran, Halls — Benchmarking Machine Learning Descriptors for Crystals. ACS Symposium Series (2022)
- Vishwakarma, Sonpal, Hachmann — Metrics for benchmarking and uncertainty quantification in ML for chemistry. Trends in Chemistry (2021)
- Afzal, Sonpal et al. — A deep neural network model for packing density predictions, applied to 1.5M organic molecules. Chemical Science (2019)
- Kang, Verma, Sonpal et al. — A Force-Kernel Reformulation of the Extended-System Adaptive Biasing Force for Free-Energy Calculations. Journal of Chemical Theory and Computation (2026)