Blog

Latest Post

Jul 20, 2024

Sub Component Embedding: Reducing Dimensionality and Improving Flexibility

VectoredIn is a tool I developed to visualize the job market and job postings from LinkedIn utilising various tools and techniques from the NLP and LLM space. To read more about the project, look here. Data For this project, I built upon this Kaggle dataset here, which originally contained over 100,000 job postings, and I increased this to just over 1 Million individual job postings. With each job posting having an average of 500 words (600 tokens) for the description alone, alongside some additional metadata, this is a reasonably large dataset. Read more

All Posts

2024

  • Jul 20, 2024 - Sub Component Embedding: Reducing Dimensionality and Improving Flexibility
  • Jun 20, 2024 - Introducing VectoredIn: A New Dimension in Job Market Analysis
  • Feb 20, 2024 - Methods for visualizing internal hidden states in Gated Recurrant networks
  • Feb 6, 2024 - Enhancing Model Performance and Stability with Multiple Loss Functions
  • 2023

  • Jun 1, 2023 - Optimizing Targets for Predictive Models
  • Jun 1, 2023 - Optimizing Targets for Predictive Models