January 2, 2026 | By Poorna Gunathilaka

What is MLIR and Why Does It Matter? Compiler infrastructure has historically been fragmented. You had one representation for the high-level language (like the TensorFlow graph) and a completely different one for the low-level machine code (like LLVM IR). Moving between these levels often resulted in lost optimization opportunities. MLIR (Multi-Level Intermediate Representation) solves this by providing a unified infrastructure. It allows compilers to define “dialects”—modular levels of abstraction that can mix and match.

June 27, 2022 | By Poorna Gunathilaka

Up to recently Javascript has been the only language which we can use to write programs for web browsers. Javascript is not suitable to run resource intensive applications such as AR-VR, video editing and 3D games. WebAssembly (wasm) was initially proposed as a compilation target for higher level languages which enabled them to run on web browsers with the additional performance gains for running computational intensive applications. Tensorflow, Unity, GoogleEarth and many more applications already use wasm to run on web browsers.

January 2, 2026 | By Poorna Gunathilaka

Consistency is often sold as the hardest part of fitness. For the last two years, I’ve mastered it. I’ve shown up to the gym more regularly than ever before in my life. But looking back, I realize that while my attendance was perfect, my effort was plateauing. I haven’t been pushing limits; I’ve been coasting in my comfort zone. So, let’s forget about the past. This post is about drawing a line in the sand and defining real targets.