Swathi
Adimulam
OCI Cloud & AI Architect at Oracle
"Smarter, Not Hotter : Green Principles for AI Design"
"Smarter, Not Hotter : Green Principles for AI Design"
Don’t miss out and join visionaries, innovators, and thought leaders from all over the world at the Women in Tech Global Conference.
Vote by Sharing
Unite 100 000 Women in Tech to Drive Change with Purpose and Impact.
Do you want to see this session? Help increase the sharing count and the session visibility. Sessions with +10 votes will be available to career ticket holders.
Please note that it might take some time until your share & vote is reflected.
Session: Smarter, Not Hotter : Green Principles for AI Design
This session invites product managers, engineers, and AI practitioners to critically examine the growing environmental cost of generative AI and large language models (LLMs)—a concern often overshadowed by excitement around innovation and productivity. As AI becomes embedded in everyday tools and workflows, I see many organizations and developers adopt LLMs by default, frequently without fully considering whether these high-energy models are necessary or sustainable for the task at hand.I will begin by grounding the audience in real-world data, such as how training a single large model like GPT-3 consumed as much electricity as an average U.S. household uses in over a century. I’ll compare this to the energy costs of traditional search and explore how billions of daily generative AI queries could equate to the energy usage of small towns.
Beyond electricity, AI infrastructure places a significant burden on our planet’s water resources—particularly through data center cooling. A striking example comes from Iowa, where Google’s data centers consumed over 20 million gallons of water in just a few months to support AI workloads—during the peak of a summer drought. As climate change intensifies and water scarcity becomes more common, it’s crucial to understand that AI doesn't just live in the cloud; it draws heavily from the ground, rivers, and reservoirs. I will highlight the growing tension between technological advancement and environmental sustainability, making a case for transparent reporting and responsible scaling of AI systems.
Through a mix of case studies, audience prompts, and practical frameworks, I’ll unpack scenarios where LLMs are often misapplied and where lighter-weight, lower-energy alternatives—such as logistic regression, decision trees, or small domain-specific models—can often offer comparable or even superior results. I’ll also demonstrate how thoughtfully engineered ontologies can simplify data interpretation and knowledge representation, reducing the need for resource-intensive models, and enabling organizations to achieve greater accuracy with smaller, more efficient AI systems.
I aim to address the risks of “AI FOMO” and clarify common misconceptions, like mistaking generation for prediction or deploying black-box models in regulated environments requiring transparency. As a strong advocate for AI-first development, I will emphasize the role of coding assistants and intelligent engineering tools—not only to boost developer productivity but also to guide sustainable AI design choices and help teams avoid unnecessary reliance on massive models.
The second half of the session will shift to actionable strategies. Attendees will learn about architectural and training-level optimizations such as model pruning, quantization, knowledge distillation, and transfer learning, as well as deployment techniques like edge computing and batch processing. I’ll discuss how these approaches can significantly reduce energy and water consumption without sacrificing performance. I will also touch on infrastructure decisions—how to choose greener data centers, right-size hardware (CPUs vs. GPUs), and implement dynamic power management for AI workloads.
To deepen engagement, attendees will participate in a guided mini-exercise, reflecting on one AI feature they’ve worked on or used and exploring how it could be made more sustainable. This will be followed by shared insights and recommendations, fostering peer learning and building a sense of shared accountability.
Participants will leave with a thorough understanding of AI’s environmental and resource impact, a checklist of best practices for green AI development, and a refreshed perspective on what it means to build responsible, ethical, and sustainable technology. This session encourages a critical yet optimistic outlook: that AI can be both powerful and planet-conscious, but only if I—and all of us—commit to building it that way.
Key Takeaways
- Increased Sustainability Awareness
- Real world usecases for using AI deterministically
- LLM Optimization techniques
- Using Coding Assistants efficiently
- Power of Ontology
Bio
Swathi Adimulam is a seasoned Cloud & AI Architect at Oracle, leading the charge in digital transformation. With over 16 years of industry experience, she architects and modernizes enterprise SaaS solutions by developing innovative, AI-powered cloud frameworks. Her expertise lies in designing bespoke solutions that solve complex business challenges through architectures that are scalable, secure, and future-ready.She is a recognized expert in Agentic AI frameworks, focusing intently on agent interoperability, seamless enterprise integration, and boosting developer productivity. A core element of her leadership is building intelligent systems grounded in robust ontological models, which significantly enhance data interoperability, contextual reasoning, and AI-driven decision-making across the enterprise.
A passionate advocate for responsible AI adoption, Swathi champions AI-first development practices, leveraging coding assistants and intelligent tools to accelerate business outcomes and empower engineering teams. She is dedicated to inclusive innovation, ensuring all AI-based solutions strictly align with ethical principles, regulatory requirements, and core business objectives.
Throughout her career, she has strategically aligned technology with organizational goals, delivering high-impact solutions that maximize value, efficiency, and compliance in a rapidly evolving digital landscape. She remains committed to harnessing ontology-driven AI to build smarter, adaptable, and trustworthy enterprise systems for the future.
