Stanford Demo — PSN Classification (GraphSAGE/GAT)
Local, reproducible pipeline across 18 protein structures with interpretability (centrality, IG).
Independent Researcher / AI Engineer
I build graph- and memory-aware agentic systems and local, reproducible AI for protein modeling and scientific infrastructure.
Stanford Demo
18-structure PSN GraphSAGE/GAT (Aug 2025)
18-structure PSN GraphSAGE/GAT (Aug 2025)
PDB2Graph
PyMOL toolkit for protein structure networks
PyMOL toolkit for protein structure networks
Local HPC
64 threads · 256 GB ECC · RTX 3090
64 threads · 256 GB ECC · RTX 3090
Projects
PDB2Graph — PyMOL Toolkit
Construct PSNs from PDB/CIF, compute centrality, export CSVs, and overlay in PyMOL.
Agentic Orchestrator v1.0 (Graph + Memory Aware)
CLI/API agent with skills: classify <protein>, plot_centrality, benchmark size_sweep.
Repo (WIP)
LRH1 Drug Screening
Notebooks + analysis for ligand screening and PSN-based interpretability.
Services
Prototype Sprint (2–3 weeks)
Dataset curation, baseline GraphSAGE/GAT, metrics/plots, reproducible run scripts.
Artifacts: repo • report • JSON logs
Local AI Stack (2 weeks)
Workstation setup (CUDA, PyG/DGL), containers, job scripts, benchmarks vs cloud.
Artifacts: infra docs • Dockerfiles • runbook
Agentic Orchestrator v1.0 (3–4 weeks)
Graph + memory aware agent (CLI/API) with demo and handoff.
Artifacts: service repo • config • demo video
About
I bridge hands-on scientific infrastructure with modern graph learning. My focus: protein structure networks, allosteric reasoning, and building reliable local tooling that labs can run and own.
Training: Stanford AI Professional Program — Machine Learning with Graphs (CS224W, in progress; demo delivered Aug 2025).
Contact
Email: david@davidfoutch.io · GitHub: davidfoutch