Environment Variables Software Architecture for Sustainability
Oct 16, 2025
Karthik Vaidhyanathan, an Assistant Professor at IIIT Hyderabad, shares his expertise on making AI development more sustainable. He discusses how AI's growth can conflict with energy goals and stresses the need for energy-efficient design. Karthik explores dynamic model selection and the importance of adapting AI systems to cleaner energy sources. He emphasizes a collaborative approach across academia and industry to integrate sustainability into software engineering, highlighting that making green software a priority can lead to cost savings and a positive environmental impact.
AI Snips
Chapters
Books
Transcript
Episode notes
Course Shifted His Research Focus
- Karthik traced his shift to sustainability after a 2017 software sustainability course in Amsterdam.
- That course changed his research focus from performance to designing environmentally mindful systems.
Apply Runtime Adaptation Lessons To AI
- Early research on adapting IoT and microservices for energy efficiency paved the way to sustainable AI work.
- Techniques for runtime adaptation in non-AI systems apply to AI pipelines too.
Treat AI As A System, Not A Single Model
- AI must be treated as an ecosystem, not just a single model, to reduce energy use and complexity.
- Design systems with sustainability as a first-class concern across the whole AI stack.
