Skip to content
View rhelmeczi's full-sized avatar
Block or Report

Block or report rhelmeczi

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
rhelmeczi/README.md

Hi there, my name is Kraig. Thanks for visiting my GitHub!

About Me

  • I'm a machine learning engineer and software developer, and I've recently completed a Master's degree in Data Science and Analytics at Toronto Metropolitan University (formerly Ryerson University).
  • I studied Physics in undergrad and finished with the highest average across all faculties in my school, earning a Governor General's Silver Medal.
  • My thesis was an investigation into the state-of-the-art few-shot learning methods for text classification. We compared the performance of Hugging Face transformer models and few-shot learning algorithms when applied to training data sets with limited labeled examples.
  • In my two year Master's degree I have first authored four publications: two in reinforcement learning and two in natural language processing using few-shot learning. You can check out my Google Scholar to view some available pre-prints/publications.

My Current Work

  • I currently work as a machine learning engineer, managing projects from conception through to deployment. In this process I handle
    • Problem conception
    • Problem elaboration
    • Model conception and design
    • Model implemention
    • Data collection, feature engineering, data labeling (including few-shot labeling)
    • Model training
    • Model deployment (i.e., ML Ops), especially via AWS
    • Monitoring of deployments
  • In addition to my machine learning work, I also act as an engineer to create user interfaces to create environments where deployed models can be tested by key stakeholders, clients, and by internal staff (often client-facing).
  • For a more formal look at my work and education history, you can checkout my LinkedIn Profile.

My Projects

I've engaged in a lot of projects over my 5+ years of development. Unfortunately, many of my worthwhile projects are contained in private repositories. While not all of them can be made public, I've been slowly extracting code that I can share with the public.

  • A subset of the code from my Master's thesis is available here. This code is used for training NLP models with access to a limited amount of training data.
  • You can checkout this repository for some of my work on using linear programming for reinforcement learning. Our work ultimately resulted in a journal publication, so I've added a small sample of the code there.

My Programming Experience

I'm extremely open to learning new systems and languages. I'm a fluent Python developer, but I have a well-established track record of learning new languages with a quick turn around:

  • I was formally trained in Java in undergrad, and learned Kotlin for my own app development project in 2020.
  • I learned Free Pascal to complete my undergraduate thesis.
  • I've been solving problems with Rust since 2020.
  • I learned C to add optimizations to my reinforcement learning projects, as the Python APIs for the linear program solver that we use are slow for our applications.
  • I learned and used NodeJS, Angular, and Javascript in a contract web development and machine learning project to introduce a front-end for clients to manually label training data for model development.

Popular repositories

  1. pattern-exploiting-training pattern-exploiting-training Public archive

    Jupyter Notebook 1 2

  2. itlp-cpomdp itlp-cpomdp Public

    Python

  3. rhelmeczi rhelmeczi Public

  4. pet pet Public

    Forked from timoschick/pet

    This repository contains the code for "Exploiting Cloze Questions for Few-Shot Text Classification and Natural Language Inference"

    Python

  5. few_shot_nlp few_shot_nlp Public

    Python

  6. website website Public