Posts

Coding LLMs - The Permanent Junior Developer

Large Language Models are perpetual junior developers.
Code review is - or at least should be - an integral part of any development team’s process. A good team should also comprise developers of varying abilities and skill sets, ranging from junior to senior, from generalist to specialist. When reviewing a pull request, as much as we desire for the process to be objective and consistent, it is often the case that it matters who the author is. [Read More]

Understanding Data Distributions in Image-Based Machine Learning

Gaining an understanding of image data distributions.
Data distributions are much less intuitive for images than other more “classical” data types, as it’s not the distribution of the actual values that are typically considered, but more of a “semantic distribution”. [Read More]

LLMs for Policy Interpretation

LLMs may be the key to making reinforcement learning policies more understandable.
Deep reinforcement learning has proven to be a highly successful method for creating intelligent systems in recent times. However, the policies - which determine the actions of an agent based on its environment – are often complex, opaque, and difficult to understand. [Read More]

Importing Vanilla JS Files

How to break up your code into multiple files when using vanilla JS.
Here’s a little tip which allows you to use ES6 import/export syntax in your vanilla JS files in supported browsers (over 95% at the time of writing). [Read More]

Deep Learning Interview

Transcript of an interview filmed for La Trobe University's Deep Learning subject
This interview was conducted by Zhen He of La Trobe University, Melbourne, for his subject CSE5DL - Deep Learning. My answers were prepared before the interview and filmed for presentation throughout the course. [Read More]