Aswin Raj Rajan
Software engineer at Microsoft. Working on Copilot for Outlook Calendar.
Hi, I’m Aswin.
I’m a Software Engineer II at Microsoft, working on Copilot features for Outlook Calendar. Before that, I helped ship the new Calendar experience in Microsoft Teams, where I focused on optimizing load time and memory usage at scale.
I studied Electrical Engineering at IIT Madras, which is where I first got pulled in by deep learning and the math underneath it. I solidified my basics through the Deep Learning Specialization from DeepLearning.AI, and recently completed the Claude Certified Architect certification.
Lately, my curiosity has shifted toward ML systems, especially how transformers and LLMs actually run under the hood, and the engineering tricks that make inference cheap and fast. Things like vLLM and PagedAttention, FlashAttention, speculative decoding, and quantization. I find the intersection of “this is a beautiful idea” and “this saves a lot of GPU time” to be the most interesting place to spend my reading hours.
This site is where I’ll keep notes, projects, and (eventually) publications as I move toward research.
If you want to chat about systems, ML, or anything you’re building, drop me an email.