NLP-powered SaaS platform for Japanese contract analysis achieving near state-of-the-art performance, later expanded to English contracts using deep learning.

Japanese contract analysis required deep linguistic understanding of complex legal terminology, sentence structures, and implicit obligations — with very limited training data for legal NLP tasks.
Combined deep learning models with domain-specific rule-based processing. Collaborated closely with legal experts for data annotation, and iterated rapidly on clause risk detection and contract information extraction models.
Transformer-based models fine-tuned for Japanese legal NLP tasks including clause classification and risk detection.
Domain-specific processing rules developed with legal experts to handle edge cases and legal conventions.
End-to-end pipeline covering data annotation, model training, evaluation, and production deployment.
Achieved near state-of-the-art performance in Japanese contract clause analysis.
Owned full ML lifecycle from data annotation through deployment.
Improved clause risk detection and contract information extraction.
Spearheaded expansion to English contract analysis using deep learning.