Hello, It's Judicaël.

Ph.D. in Natural Language Processing &
Machine Learning

Master in Data Science
Bachelor in Computer Science

Freelancer in AI/NLP/Data Science and Software Engineering

Do you need help with your projects?

Linkedin : here | My Blog : here My CV : here

GSM : +32 499 15 23 95

Mail : pro.judicael.poumay@gmail.com

Available in the following cities : Liège, Namur, Brussel, Leuven, Maastricht, Aachen.

My incredible clients & projects

Developping & operationalizing AI tools for customer and internal use through the use of generative models such as Mistral and GPT. Tools used include AzureML, PromptFlow, docker and kubernetes. The projects are still experimental and ongoing, I can’t provide more information at this point. 

Since : September 2024

Performed Research in NLP to develop a novel algorithm and measure for Complex Terminology Extraction using Python, Scipy, scikit-learn and NLTK to help with Automatic thesaurus learning. Automatic thesaurus learning is the process of automatically creating a repository of concepts and their definitions from a set of documents. The learned thesaurus can then be used by readers to make their life easier. In particular, it aims to help them understand heavily technical texts such as financial regulatory documents published by DG FISMA.

Lasted : 8 Months

Performed Research in NLP including LLM, Topic models, Word Embeddings and Evaluation methods. I published 4 papers in 4 years and I am proud to say that one of my papers  was published in EMNLP2021, the 2nd best conference in NLP : “A Comprehensive Comparison of Word Embeddings in Event & Entity Coreference Resolution.” During this period I have also used other tools for research including Python, SQL, AWS EC2, PyTorch, Scipy, NLTK, Scipy, scikit-learnTensorFlow, spacy and HuggingFace models, Javascript, HTML, CSS.

Lasted : 4 Years

Performed Research in IT job market to discover trends and actionable insights. Through NLP methods such as topic models, anomaly detection and NER, we were able to draw a picture of the evolution of job offers over the last decade. This research demonstrated to Orange the feasibility of a larger  project of strategic market watch. 

Lasted : 4 Months

Developed stochastic models for Trend Analysis using LLMs and Topic Modelling in Python. Using MongoDB, Flask and Docker I produced an MVP with an easy to use web interface. With BeautifulSoup and Selenium I crafted crawlers to retrieve web data to train my models. During this period I have also used other tools for research and production including SQL, AWS EC2, PyTorch, Scipy, NLTK, Scipy, scikit-learnTensorFlow, spacy and HuggingFace models, Javascript, HTML, CSS.

Lasted 5 Years

I have developed and still maintain LLM based Chatbots for coaching and psychological therapy using Python, LangChain and the OpenAI API. Those Chatbots have been designed using an underlying Finite-State Machine. Designing these chatbots required Complex Prompt Engineering methods as well as hallucination mitigations techniques. Additionally, we are also doing research on the explainability of LLM through statistical analysis.

Since: 2023

I Teach AI in video games and transmedia to bachelor and master students. Through my lectures I aim to demystify Artificial Intelligence and related concepts such as Prompt Engineering and Ethics in AI. I also Teach programming in C# for AI Behavior in Games with Unity.

Since : 2023

CrazySquare, a small reflex Game for Android phone made in JAVA using the Android SDK. 

Released in 2016.

No longer Available on the Android PlayStore since 2022.

I have designed website for Indah (a non-profit) on a volunteering basis using  PhP, Javascript/Jquery, CSS and HTML. 

Lasted : 5 ans

I have designed and still maintain a website for Engemab using WordPress, Divi, PhP, Javascript/Jquery, CSS and HTML. 

Since : 2023

The Thought Process (thethoughtprocess.xyz) My own popular Science blog made with WordPress about AI/LLM/NLP and other subject that interest me

Since : 2019

Mes publications

  • Using meaning instead of words to track topics (NLDB 2022)
  • A Comprehensive Comparison of Word Embeddings in Event & Entity Coreference Resolution (EMNLP 2021)
  • HTMOT: Hierarchical Topic Modelling Over Time (RANLP 2023)
  • Evaluating Unsupervised Hierarchical Topic Models Using a Labeled Dataset (RANLP 2023)
  • NLP Methods for Weak Signals Detection from Unstructured Text (DOCTORAL THESIS)
  • For more see my Google Scholar Page

Vous aussi vous avez besoin d’aide sur l’un de vos projets?

Contactez moi !

Linkedin : here | My Blog : here

GSM : +32 499 15 23 95

Mail : pro.judicael.poumay@gmail.com

Available in the following cities : Liège, Namur, Brussel, Leuven, Maastricht, Aachen.