Antreas Antoniou

Research Associate in Machine Learning, interested in the intersection of multi-modal learning and meta-learning.

I am currently a Senior ML Research Scientist at Pieces for Developers, where I lead and supervise the ML team’s research and development efforts. My responsibilities include setting and managing the research agenda, overseeing the development of micro-models and foundation models, and ensuring their efficient deployment across diverse hardware platforms—including CPUs, GPUs, NPUs, and LPUs—with sub-second inference latency. I maintain direct involvement across all aspects of machine learning, from conceptualization and experimentation to optimization and deployment.

Pieces is the universal context memory butler, automatically capturing and organizing your digital life entirely on-device, ensuring complete privacy without external data sharing. It constructs a rich, personalized knowledge base, structuring your activities into intuitive narratives. Additionally, through an MCP server, Pieces provides precise context to enhance interactions with your chosen copilot or AI tools, delivering insightful and contextually relevant responses. By directly observing your screen, Pieces elevates your productivity and insights to unprecedented levels.

Previously, I served as the Lead Research Scientist at Malted AI, focusing on efficient LLM training, distillation, synthetic data generation, and integrating LLM-as-a-Judge systems for robust annotation, decision-making, and reasoning. Before that, I was a Research Associate in Machine Learning at the University of Edinburgh, supervised by Prof. Amos Storkey, affiliated with the BayesWatch research group and the Adaptive and Neural Computation (ANC) research institute.

Overarching Theme

My overarching goal is to emulate the trajectory of human-like representation learning, starting from foundational representations akin to those seen in human infants but without requiring eons of evolutionary fine-tuning. This serves as a springboard for my broader research ambition: to investigate how these infant-like representations can be fine-tuned in concert with higher-level abstract concepts to pave the way for general artificial intelligence systems. To achieve this, my research focuses on scalable, data-efficient, and generalizable self-supervised learning techniques in a multi-modal setting. I integrate insights from neuroscience and evolutionary computation to explore optimal learning sequences and curricula, paying close attention to architectural choices and their corresponding training recipes.

Research Focus

Leading my research interests is Multi-Modal Learning, specifically targeting the synergistic integration of text, images, audio, and video data. This is followed by the development of Self-Supervised Methods, inspired by mechanisms of infant learning and principles of evolutionary computation.

Additional key areas include Meta-Learning, Adversarial Learning, and Optimization Techniques Inspired by Evolutionary Optimization. These are applied across both differentiable and gradient-free optimization paradigms. Other relevant research dimensions include Inductive Biases, Scalability, Computational Efficiency, and Memory-Augmented Neural Networks.

Research Philosophy

Operating within a pragmatic framework, my research aims to identify high-leverage focal points conducive to in-depth investigation. This allows for the efficient allocation of both computational and cognitive resources. In line with both evolutionary tenets and the Pareto Principle, my methodology can be seen as focusing on the “fittest” 20% of research avenues likely to contribute 80% of impactful results.