Process & timeline
1) Scope & success metrics
Define the biological or materials question, constraints, and “publishable” endpoints.
2) Design
Draft methods (docking/MD/NEB/surface), compute plan, and validation criteria.
3) Build & run
Spin on GPU/cluster with checkpoints; keep runs auditable and restart-safe.
4) Analyze
Compute landscapes, clusters, feature importance; generate figures and tables.
5) Deliver
Provide report, data bundle, and reusable pipeline with how-to.
Common pitfalls avoided
- Under-spec’d endpoints → define success metrics early
- Untracked parameters → config & logs for full provenance
- Black-box ML → interpretable feature importance for trust
- Non-reproducible runs → scripted pipelines & versioning