My name is Hafedh Hichri, I also go by not-lain in the wired, I'm a Software Engineer at Feyn (previously called Chonkie), a Hugging Face Fellow and a Qdrant Star.
I'm passionate about making AI accessible through open-source contributions to libraries like transformers, unsloth, peft, .... I studied Computer Science at the National School of Electronics and Telecommunications of Sfax (ENET'Com).
blogs
[6]
- I write about machine learning, natural language processing, and open-source software mostly in HuggingFace. Here are some of my recent blog posts:
- Visualizing How VLMs Work
- KV Caching Explained: Optimizing Transformer Inference Efficiency
- Mastering Tensor Dimensions in Transformers
- PyTorchModelHubMixin: Bridging the Gap for Custom AI Models on Hugging Face
- RAG using huggingface tools
- Image-based search engine
open source
contributions[8]
- Most of my work comes in this field the form of contributions to other libraries. A couple of notable libraries I contributed to are:transformersState-of-the-art AI library by Hugging Face.unslothEfficient fine-tuning for LLMs.peftParameter-Efficient Fine-Tuning methods for large models by Hugging Face.
gradioPython Library for building machine learning web-applications.chonkiemaintainerAI library for efficient data chunking and building RAG pipelines.* Currently working at this company and serving as their maintainer.huggingface.jsJavaScript client for Hugging Face APIs. I contributed to API features and bug fixes.
falSDK Client for FALargillaCollaboration tool for AI engineers and domain experts to build high-quality datasets.
- Most of my work comes in this field the form of contributions to other libraries. A couple of notable libraries I contributed to are:
maintained[2]
- I also maintain a couple of libraries:
models @ feyn[2]
- I also created these models at my company, Feyn:
integrations[8]
- I also contributed to integrating several models with Hugging Face using PyTorchModelHubMixin, including:
birefnetSOTA Background Removal model for camouflaged image segmentation (chameleon, etc.).ben2SOTA Background Removal model for detailed image segmentation (hair strands, etc.).
matanyoneVideo propagation model for consistent object segmentation across frames.
matanyone2Video propagation model for consistent object segmentation across frames.
anime-segAnime image segmentation model.
araclipArabic CLIP model for image-text tasks.
rmbg-1.4Background removal model.
swarmformerTransformer variant using hierarchical local-global attention reducing cost with strong accuracy.
- I also contributed to integrating several models with Hugging Face using PyTorchModelHubMixin, including:
now
currently
learning
exploring
thinking about
talks & news
[8]
- I was a judge at IndabaX Tunisia and gave a talk about best practices in finetuning [slides]
- I shared my knowledge about Chonkie and Qdrant integration in discord [slides]
- I was invited to Data & Beyond to talk about RAG and different approaches in the field [slides]
- I gave a keynote at Vectors in Orbit about Chonkie
- I gave a speech at PyData about AI agents [slides]
- I gave a talk at the Hugging Face server about Building how VLMs work [slides]
- I was one of the judges at Artificial Intelligence National Summit V3 and I gave a talk about Chonkie [slides]
- I gave a talk at the Genesis Labs-INSAT about AI in Healthcare [slides]