About This Notebook

I'm a Machine Learning Engineer with a PhD in Computer Vision. My interests lie in model development, GPU optimization, and efficient inference.

My background combines machine learning research with software engineering. I've worked on clinical computer vision systems, model optimization, and inference runtimes, often with constraints involving computational resources, latency, or data availability.

This website is where I document that work.

Why This Notebook Exists

I wanted a place to keep a record of my technical projects beyond their source code.

A GitHub repository shows what was implemented, but it doesn't always explain why certain decisions were made, which alternatives were considered, or what the experiments revealed.

Here, I try to document those aspects alongside the implementation.

Some projects begin with a specific performance target. Others start because I want to understand an architecture, experiment with an unfamiliar technique, or investigate a technical question.

Not every experiment produces the expected result.

For example, HotStart was an attempt to reduce model startup latency by overlapping weight loading and inference. The implementation ended up being slower than conventional loading.

I consider that worth documenting too. Understanding why an approach failed can be as useful as demonstrating one that worked.

What You'll Find Here

The website is organized into three main sections:

• Work — Projects, implementations, benchmarks, and engineering experiments.
• Writing — More detailed technical investigations, including implementation decisions and experimental results.
• Research — Scientific contributions, doctoral research, and publications.

The goal is to maintain a record of what I've built, what I've measured, and what I've learned along the way.