| AI agents in software engineering | AI-powered software tools that automate or semi-automate tasks throughout the software development lifecycle, from requirements gathering to deployment. |
| Agentic AI | Agentic AI refers to autonomous or semi-autonomous software systems that use advanced AI techniques to perceive, make decisions, take actions, and achieve goals in their digital environments, often adapting to changing conditions. |
| GraphRAG | GraphRAG is a technique based on knowledge graph to improve the accuracy, reliability and explainability of retrieval-augmented generation (RAG) systems. |
| LLM Fine-Tuning | LLM fine-tuning is the process of adapting a pre-trained model for domain-specific, task-specific needs. |
| Generative Engine Optimization (GEO) | Optimize content for AI powered search engines and generative models, so that it aligns with how generative AI interprets user queries. Strategy to boost visibility in AI-driven search engines such as Perplexity, Google AI Overviews. |
| World models | Inspired from the mental models of the world that humans develop naturally. Mental models help us make sense of the world and make decisions. Simulate real world environment to allow AI to learn in an human-like way. |
| Autonomous agents for migration | Advanced AI systems designed to operate independently, without constant human oversight, to achieve specific goals. |
| AI Agent Development Frameworks | Accelerate the creation of AI agents and AI-powered applications that can autonomously perform complex tasks; provide an abstraction layer to enable prompt chaining, model chaining, interfacing with external APIs, retrieving contextual data from data sources and maintaining statefulness |
| Explainable AI | The objective of explainable AI (XAI) is to provide an explanation for the decisions of a model, by trying to map its inputs and outputs in order to understand the underlying rationale that lead the model’s final decisions. |
| Graph ML | Graph Machine Learning applies machine learning techniques to graph structures to discover, predict, and analyze complex relationships between data. There are mainly two approaches : ‘featurization’, where which metrics calculated using graph analytics tools serve as input for a machine learning model, and ‘native Graph ML’, with new ML techniques specifically designed to make optimal use of graph structures. |
| Legacy Code & AI | Legacy code is a problem for many organisations because of its high maintenance costs. With the advent of LLMs and coding assistants, we must find out if these smarter tools can be leveraged to better understand or modernize legacy projects. |
| AI agent evaluation | AI agents are hard to trust and improve because their behavior is complex and unpredictable, making traditional evaluation methods inadequate. Specific evaluation methods and tools are needed to measure their performance, reliability, and safety. |
| Agentic RAG | An approach that extends classical RAG with agentic capabilities, where the quality of the retrieval and generation steps in the RAG process is iteratively evaluated and adjusted. |
| Speech Interfaces | Speech interfaces offer real-time, highly accurate speech recognition and speech synthesis, with robust support for multiple languages. They feature low latency and allow users to interrupt responses, ensuring a smooth, efficient, and user-friendly interaction experience. |
| LLMs for code | Use Large Language Models to support software development by generating code, suggesting improvements, and helping diagnose errors. This can increase developer productivity, improve code quality, and accelerate software delivery and maintenance. |
| Natural Language Processing | Computational understanding, analysis and generation of human language across text and speech. Nowadays with a focus on multilingual communication and specialized language models (jargon, dialects) for improved accessibility and service quality. |
| Human-AI Interaction | Design AI systems so that the abilities of both human and AI systems are combined in an efficient and trustworthy way. |
| Neuro-Symbolic AI | Neuro-symbolic AI is a form of composite AI that combines machine learning methods and symbolic systems to create more robust and trustworthy AI models. . |
| Federated Machine Learning | Federated learning is a distributed machine learning approach where a model is trained on decentralized data located on various devices. |
| Knowledge/Model Distillation | The process of transferring knowledge from a large model to a smaller one (without loss of validity). Not to be confused with model compression. Aims to retain domain-specific knowledge, minimizing accuracy loss. Can turn a Teacher model into a Student model. |
| Multimodal AI | Multimodal AI refers to AI systems that can process and understand multiple types of data, such as text, images, audio, and video, simultaneously. This enables more comprehensive and context-aware interactions. |
| Humanoid Robots | Humanoid robots represent a convergence of advanced mechanics, artificial intelligence, and human-centered design, creating machines that can navigate human environments and interact naturally with people. |
| Self Supervised Learning | Self-supervised learning (SSL) allows to train a model on unlabeled data, by generating a supervisory signal from the data themselves. It is often used for representation learning, or as a pre-training before a supervised learning phase. With GraphSSL, it finds interesting applications in graph mining, for instance for graph embedding or unsupervised anomaly detection. |