I am a PhD candidate at Dalhousie University, working with Dr. Tushar Sharma at the SMART Lab.
Here’s the thing about software: we’ve spent decades adding layers of abstraction so that humans can write and understand code more easily. Makes sense, it worked. But every layer we added also moved software further from the hardware it actually runs on, and that distance costs energy. A lot of it, quietly.
Now AI is changing the equation. When even non-developers can write code with AI tools, do we still need all those layers that were built for human convenience? I don’t think so. Strip some of them away, let software and hardware talk more directly, and you get something closer to resonance, where the two actually work together instead of around each other. That’s where the energy savings are.
So that’s what I work on. CodeGreen lets you measure energy consumption at the function level across platforms (Linux, macOS, Windows, GPUs). FlipFlop uses static analysis to optimize CUDA kernels for energy. Greenlight is a dataset of 1,284 TensorFlow API calls with energy profiles. Basically, if you want to know where your code is wasting energy, these tools can tell you.
Before the PhD, I was a Software Engineer at Fidelity Investments working on the WealthScape platform (2020-2022). Did my B.Tech in Computer Science at Visvesvaraya National Institute of Technology, India.
I’m looking for research and engineering roles where software meets hardware and energy efficiency matters. If that sounds interesting, here’s my CV, or just reach out.