julesbelveze Goto Github PK

🤓 Jules Belveze
┣━━ 📦 Open Source
┃   ┣━━ tsa                                - Dual-attention autoencoder
┃   ┣━━ bert-squeeze                       - Speed up Transformer models
┃   ┣━━ bundler                            - Learn from your data
┃   ┣━━ nhelper                            - Behavioral testing
┃   ┗━━ time-series-dataset                - Dataset utilities
┣━━ 👍 Contributions
┃   ┣━━ 🤗 Hugging Face Ecosystem
┃   ┃   ┣━━ t5-small-headline-generation   - t5 for headline generation
┃   ┃   ┗━━ tldr_news                      - Summarization dataset
┃   ┣━━ ❄️ John Snow Labs Ecosystem
┃   ┃   ┗━━ langtest                        - Deliver safe & effective NLP models
┃   ┣━━ 💫 SpaCy Ecosystem
┃   ┃   ┗━━ concepCy                       - SpaCy wrapper for ConceptNet
┃   ┣━━ bulk                               - contributed the color feature
┃   ┗━━ FastBERT                           - contributed the batching inference
┗━━ 📄 Blogs & Papers
    ┣━━ Atlastic Reputation AI: Four Years of Advancing and Applying a SOTA NLP Classifier
    ┣━━ Real-World MLOps Examples: Model Development in Hypefactors
    ┣━━ LangTest: Unveiling & Fixing Biases with End-to-End NLP Pipelines
    ┣━━ Case Study: MLOps for NLP-powered Media Intelligence using Metaflow
    ┗━━ Scaling-up PyTorch inference: Serving billions of daily NLP inferences with ONNX Runtime

I currently work as a MLOps/ML Engineer at @Ava where I lead the AI efforts.

My previous experiences include working on several multilingual natural language processing tasks such as sentiment analysis, NER, topic modeling, summarization, semantic search, ...

I believe that automating model development and deployment using MLOps enables faster feature releases. To achieve this goal, I have worked with various tools such as PyTorch Lightning, Hydra, Neptune.ai, ONNXruntime, Metaflow, Label Studio, among others.

Apart from this, I have worked extensively with Deep Learning and Time Series and completed my Master's Thesis on Anomaly Detection in High Dimensional Time Series. Additionally, I am keenly interested in exploring state-of-the-art techniques to speed up the inference of Deep Learning models, especially Transformer-based models.