Collected molecules will appear here. Add from search or explore.
Training and inference of Decision Tree models directly on encrypted data using Fully Homomorphic Encryption (FHE).
Defensibility
stars
29
forks
4
The project is a three-year-old research artifact from Intuit's engineering team. While the underlying mathematics of FHE (Fully Homomorphic Encryption) are complex and provide a theoretical technical moat, this specific repository lacks the traction or maintenance to be considered a viable product or infrastructure. With only 29 stars and zero recent activity (velocity 0.0/hr), it represents a 'fire-and-forget' academic release rather than a developing ecosystem. In the time since this was released, specialized FHE startups like Zama (with 'concrete-ml') and OpenMined have produced much more robust, performant, and better-documented libraries that support a wider range of ML models (XGBoost, Random Forests) with better compilers. The platform domination risk is medium because while frontier labs (OpenAI/Anthropic) don't care about classical ML over FHE, infrastructure giants like Microsoft (who owns SEAL) and Google (who has their own FHE transpiler) are the natural owners of this layer. The displacement horizon is very short (under 6 months) because anyone looking to actually implement this today would choose Zama's Concrete-ML over a stagnant reference repo.
TECH STACK
INTEGRATION
reference_implementation
READINESS