While Artificial Intelligence (AI) is increasingly used to support student learning, its potential to empower educators, particularly in the Indian context, has been underexplored. Existing approaches often position teachers as passive users of AI tools rather than as active creators. This paper advocates an educators as developers approach, emphasising the role of faculty in designing AI-enhanced educational interventions. We explore how locally installed Small Language Models (SLMs) - sustainable, modifiable, transparent and accessible AI systems, can serve as cognitive scaffolds to support computer science (CS) faculty to create software products. To test this approach, we conducted a one-day, hands-on workshop with 22 CS faculty members. Participants used SLMs and rapid prototyping techniques to build functional educational tools. Post-workshop surveys indicated increased confidence, technical proficiency, and readiness to apply the tools in classroom settings. Over 90% of participants rated the experience positively. This experience report details the design, outcomes, and participant feedback, offering practical insights and recommendations for replications. To the best of our knowledge, this is the first study to document the use of locally hosted SLMs as scaffolds for educator-led AI development, particularly in the Global South.