Hi there 👋

My name is Niels, I'm 28 years old and I live in Belgium.

I'm currently working as an ML engineer @ 🤗 HuggingFace, where I'm part of the Open-Source team.

I work on HuggingFace Transformers, a Python library implementing several state-of-the-art AI algorithms, all based on the original Transformer by Google.

I love making AI more accessible to anyone. So far, I've contributed the following algorithms to HuggingFace Transformers:

• TAPAS, by Google AI
• ViT, by Google AI
• DEiT, by Facebook AI
• DETR, by Facebook AI
• BEiT, by Microsoft Research
• CANINE, by Google AI
• LUKE, by Studio Ousia
• LayoutLMv2 and LayoutXLM, by Microsoft Research
• DINO, by Facebook AI
• TrOCR, by Microsoft Research
• SegFormer, by NVIDIA
• ImageGPT, by OpenAI
• Perceiver/Perceiver IO, by Deepmind
• MAE, by Facebook AI
• ViLT, by NAVER AI Lab
• ConvNeXT, by Facebook AI
• DiT, by Microsoft Research
• GLPN, by KAIST (Korea Advanced Institute of Science and Technology)
• DPT, by Intel Labs
• YOLOS, by School of EIC, Huazhong University of Science & Technology
• TAPEX by Microsoft Research
• LayoutLMv3, by Microsoft Research
• VideoMAE, by Multimedia Computing Group, Nanjing University
• Donut, by NAVER AI Lab
• X-CLIP, by Microsoft Research
• Deformable DETR, by SenseTime Research
• MarkupLM, by Microsoft Research
• LiLT, South China University of Technology
• Table Transformer, by Microsoft Research
• CLIPSeg, by University of Göttingen
• Audio Spectrogram Transformer, by MIT Computer Science and Artificial Intelligence Laboratory, Cambridge
• BiT, by Google AI
• ViT Hybrid, by Google AI
• Swin2SR, by CAIDAS, University of Würzburg
• GIT, by Microsoft Research
• UPerNet, by Peking University
• BLIP-2, by Salesforce
• InstructBLIP, by Salesforce
• FocalNet, by Microsoft Research
• PerSAM, by Shanghai Artificial Intelligence Laboratory
• DINOv2, by Meta AI
• ViTMatte, by HustVL
• Nougat, by Meta AI
• OWLv2, by Google AI
• SigLIP, by Google AI
• SlimSAM, by National University of Singapore
• Depth Anything, by The University of Hong Kong/TikTok
• UDOP, by Microsoft Research
• ZoeDepth, by Intel Research
• DINOv2 with Registers, by Meta
• ViTPose, by University of Sydney
• MetaCLIP 2, by Meta AI

Besides that, I help others add models to the library, including:

• Swin Transformer, by Microsoft Research
• mLUKE, by Studio Ousia
• Nyströmformer, by University of Wisconsin-Madison
• YOSO, by University of Wisconsin-Madison
• PoolFormer, by Sea AI Lab
• CvT, by Microsoft Research
• GroupViT, by NVIDIA
• TextNet, by Nanjing University
• Grounding DINO, by Tsinghua University
• KOSMOS 2.5, by Microsoft Research

I'm mostly working with PyTorch. For IDEs, I work with Visual Studio Code and Google Colab. It's all you need.

I learned everything about deep learning through self-study, mainly thanks to Andrej Karpathy's cs231n course at Stanford University (lectures are free on Youtube and assigment solutions can be found on Github).