Back to browse
GitHub Repository

json2vec turns nested, ragged data into neural representations. It lets users define typed schemas: numbers, categories, sets, dates, entities, text, etc., then trains models for prediction and embeddings. The library supports MLM-based pretraining workflows, mutations, and serving. It is built for data that does not fit cleanly into flat tables.

12 starsPython

Json2vec – build/train/deploy models with nested data structures

by granthamctaylor·Jun 1, 2026·3 points·0 comments

AI Analysis

●●SolidBig BrainNiche Gem

Schema becomes model architecture—no manual feature engineering for nested data.

Strengths
  • Schema-as-architecture eliminates manual feature engineering for nested structures
  • Lightning integration means standard training loops work out of the box
  • Embeddings keyed to schema addresses preserve semantic structure in outputs
Weaknesses
  • ML infrastructure is crowded; needs proof it beats Featuretools or tabular transformers
  • Only 8 stars suggests early stage—production readiness unproven
Category
Target Audience

ML engineers working with nested structured data

Similar To

Featuretools · H2O AutoML · PyTorch Tabular

Similar Projects

AI/ML●●Solid

BNNR – a closed-loop pipeline for improving vision models

XAI-driven model improvement loop, but Weights & Biases already tracks experiments better.

Big BrainNiche Gem
dominka
102mo ago