Our members are building the future of AI

An efficient public implementation of tools for deepfake image and video detection.

A blazingly fast, scalable, open source communication, data replication and collective computing library for end-to-end zero copy, hardware accelerated data sharing, replication, or cooperative parallel computing.

The Do Not Answer project provides an open-source English language dataset to evaluate LLMs' safety mechanism at a low cost. The dataset is curated and filtered to consist only of prompts to which responsible language models do not answer. Besides human annotations, Do not answer also implements model-based evaluation, where a 600M fine-tuned BERT-like evaluator achieves comparable results with human and GPT-4.

Docling transforms PDF documents into rich JSON or Markdown formats with ease and speed, making it the perfect companion for your knowledge engineering project, feeding hungry LLMs with high quality training data or providing rich input to RAG.

Robotic support systems for elderly both from a care taker and a care giver perspective. A focus is on embodied AI to advance manipulation capabilities.

Fast-LLM is an innovative open-source library that prioritizes speed, flexibility, and convenience to significantly accelerate training your LLM.

Lightning AI has been maintaining key open source frameworks widely used today in GenAI and general deep learning. PyTorch Lightning, Lightning Fabric and TorchMetrics are powering a portion of the GenAI ecosystem today (e.g. StableDiffusion, NeMO, TinyLlama).

Free education delivered through videos and open source educational resources (notebooks, projects) and free compute credits.

An end-to-end trained open-source LMM which provides visual grounding capabilities with the flexibility to process both image and region inputs. This enables the new unified task of Grounded Conversation Generation that combines phrase grounding, referring expression segmentation and vision-language conversations.

LLM-powered Data Augmentation for Enhanced Crosslingual Performance in commonsense reasoning datasets, where the available training data is extremely limited.

A report evaluating the feasibility, benefits, and limitations of using generative AI technologies in an educational setting and its impact on learning outcomes.

Techniques to find molecules with desirable properties without the need of extensive research and practical work.