Summary：**jax-cddd added to PyPI****Introduction** The cheminformatics community welcomed a new tool this w

referrerpolicy="no-referrer"
style="max-width:100%;height:auto;display:block;margin:0 auto;">

**jax-cddd added to PyPI**

**Introduction**
The cheminformatics community welcomed a new tool this week as the JAX‑based implementation of Continuous Data‑Driven Descriptors (CDDD) landed on the Python Package Index. Dubbed *jax‑cddd*, the library enables researchers to embed and reconstruct molecular SMILES strings using the speed and automatic differentiation capabilities of JAX. By bringing CDDD to a widely accessible package manager, the project lowers the barrier for integrating sophisticated descriptor generation into deep‑learning pipelines for drug discovery, materials design, and virtual screening.

**Key Developments**
The release bundles several notable upgrades over the original TensorFlow version. First, the core encoder‑decoder architecture has been rewritten in JAX, allowing seamless GPU and TPU acceleration without sacrificing the deterministic, invertible mapping that CDDD is known for. Second, the package includes pre‑trained weights for the ZINC‑250k dataset, so users can immediately generate 100‑dimensional continuous descriptors for any SMILES input. Third, a lightweight API mirrors the popular *rdkit* interface, making it easy to swap in *jax‑cddd* for existing workflows that rely on Morgan fingerprints or MACCS keys. Unit tests and a comprehensive README ensure reproducibility, while the MIT‑licensed code encourages both academic and commercial adoption.

**Industry Analysis**
The timing of this release aligns with a surge in interest for differentiable chemistry representations. As more companies adopt end‑to‑end learnable pipelines—where molecular generation, property