Master’s student specializing in Artificial Intelligence at the University of Essex, supervised by Professor Luca Citi.

With over three years of experience developing deep learning pipelines, I have contributed to the open-source community and engaged in research collaborations. My ultimate objective is to study the cognitive mechanisms underlying intelligence and develop agents capable of reasoning and interacting with the real world.

My research focuses on overcoming the challenges faced by current AI models, particularly in reasoning and decision-making in complex environments. I explore fully differentiable approaches for multi-step reasoning in LLMs, decision-making, and zero-shot learning within uncertain environments. Key areas of interest include:

• Developing new architectures for coherent multi-step inference
• Transformers and attention mechanisms
• Generative models, multimodal learning, and self-supervised learning
• Creating specialized networks for memory, goal-directed planning, spatial reasoning, and error detection and conflict monitoring

I have developed and maintained a number of Python libraries and standalone projects. Some of my major projects include:

• TransformerX

  

  A Python library for building transformer-based models with multiple building blocks and layers needed for model creation. Currently supports TensorFlow, with PyTorch and JAX support coming soon.

• Emgraph

  A Python library for developing, training, and evaluating knowledge graph representation learning. It includes a small model zoo for benchmarking and comparing new models.

• Bigraph

  Provides an easy-to-use API for working with bi-partite graphs, addressing the complexities of applying standard graph algorithms. Supports GPU computation with CUDA and graphic drivers.

• TASE

  A lightning-fast audio full-text search engine on top of Telegram. It allows users to quickly find relevant high-quality audio files without navigating through numerous irrelevant channels.