This research introduces the Shakti Small Language Models (100M-500M parameters) designed specifically for edge devices with limited computational resources.

• Domain-specific optimization across medical, legal, security, and engineering fields
• Reduced resource requirements through efficient architectures and quantization techniques
• On-device processing that enhances privacy and reduces cloud dependency
• Practical deployment on smartphones, smart appliances, and IoT systems

Medical Impact: Enables secure, private processing of sensitive healthcare data directly on medical devices, potentially allowing for real-time clinical decision support without compromising patient privacy or requiring constant connectivity.

Fine-Tuning Small Language Models for Domain-Specific AI: An Edge AI Perspective