Synergos
Synergos is AISG's open-source Federated Learning framework, enabling multiple organizations to jointly train ML models without sharing raw data. It is a core component of AISG's "privacy-preserving AI" toolchain, designed to align with PDPA compliance needs.
📖 What it is
Synergos provides:
- Horizontal federated learning: multiple organizations with the same features but different samples (e.g., several banks sharing the same model schema)
- Vertical federated learning: multiple organizations with the same samples but different features (e.g., a bank partnering with a telco)
- Encrypted communication: gradients and other intermediate results are transmitted under encryption during training
- Visual interface: lets non-technical users configure federated-learning experiments
Use cases: joint anti-fraud modeling in finance, multi-hospital medical research collaboration, cross-enterprise data partnerships.
🤖 Relation to AI
In privacy-preserving AI, Synergos is Singapore's flagship open-source tool. Federated learning isn't a new concept, but mature, usable open-source frameworks are scarce — Google's TFF, FATE, and others all have their limitations. Synergos strikes a balance between ease of use and privacy guarantees.
That said, commercializing federated learning is hard everywhere in the world — beautiful in theory, but real deployments hit a wall of engineering and organizational coordination problems. Synergos has limited verified industrial-deployment data to date.
🇸🇬 Relation to Singapore
Synergos is an important tool for AI data compliance in the PDPA era — enabling joint AI work while keeping data within local jurisdictions.
Across the seven transmission levers:
- Lever 3 (Industry Adoption): privacy infrastructure for cross-organization data collaboration
- Lever 4 (Governance): aligned with PDPC data protection requirements
Take: Synergos is one of AISG's "frontier bets" — solid technically, slow to land commercially, but representing the global "privacy-preserving + AI" direction.