Tirtharaj Dash | interests

Research Interests

My research focuses on developing deep learning methods that combine neural learning and reasoning with symbolic learning and reasoning, with applications in computational biology and drug discovery.

Neurosymbolic AI

I develop methods that integrate deep neural networks with symbolic knowledge representations, particularly Inductive Logic Programming (ILP). Neural networks learn well from large datasets but struggle with small data and lack interpretability. Symbolic methods like ILP are data-efficient and interpretable but cannot handle raw data well. My work combines both approaches to build models that are accurate, interpretable, and data-efficient. This includes constructing neural networks from relational knowledge and extracting symbolic rules from trained networks. My key contributions include developing Compositional Relational Machines (CRMs) and methods for including domain knowledge into graph neural networks. See my NeSy repository for implementations.

Graph Representation Learning

I work on neural networks for graph-structured data, focusing on designing better graph convolution and pooling operations. My research addresses how to incorporate background knowledge into GNNs to improve their performance and interpretability. Applications include molecular property prediction, drug discovery, and biological network analysis. I have developed several methods including BotGNN (knowledge inclusion via mode-directed inverse entailment) and VEGNN (vertex-enriched graph neural networks). Further, I am very motivated to involve myself in the evolving areas of “Geometric Deep Learning”. If you are new to the world of GNNs, you may want to go through these introductory slides on GNNs.

Computational Biology and Genomics

I apply machine learning to problems in genomics and life sciences. My recent work includes developing foundation models and deep sequence models for studying gene regulation, predicting gene expression from DNA sequences to understand cis-regulatory logic, and building genomic foundation models to understand codon usage in eukaryotes. I also work on multi-omics analysis for cancer biomarker discovery, using Boolean analysis and machine learning to study macrophage polarisation in cancer, and analysing single-cell and spatial transcriptomics data in the context of cancer. These works combine large- and small-scale genomic and transcriptomic datasets with interpretable AI methods to extract biological insights.

Drug Discovery

I develop AI methods for early-stage drug design, including using neuro-symbolic models to generate novel drug candidates with desired properties, molecular property prediction using graph neural networks, incorporating chemical domain knowledge into deep learning models, and developing explainable models for compound screening. My approaches emphasise combining domain expertise from chemistry and biology with data-driven learning.

Deep Learning Foundations

I work on fundamental aspects of deep learning, including large language models to build reliable AI systems for scientific applications. This includes developing methods for model calibration and uncertainty quantification to ensure predictions are trustworthy. I also work on knowledge distillation and model compression to make large models practical for deployment. My research explores transfer learning techniques that allow models trained on one scientific problem to generalise to related problems. Additionally, I focus on explainability and interpretability methods to understand how neural networks make decisions, which is critical for scientific discovery and validation.

Prospective Students

I am looking for motivated PhD and Master’s students interested in AI for science, particularly at the intersection of machine learning and biology. Strong programming skills in Python and PyTorch are valuable, along with a background in either AI/ML or computational biology. If you are interested in working on neurosymbolic AI, graph neural networks, or applications in genomics and drug discovery, please contact me (via email) to discuss potential research opportunities.

Research Consultancy

I am available for research consultancy with industry partners working on AI applications in life sciences, healthcare, and drug discovery. My expertise includes developing custom machine learning solutions that incorporate domain knowledge, building interpretable AI models for scientific applications, and designing graph neural networks for molecular and biological data. I have prior experience collaborating with industry through projects with TCS Research and Reflexis Systems. If your organisation is interested in leveraging neurosymbolic AI or explainable deep learning for scientific or business problems, please reach out to discuss potential collaborations.

Teaching Interests

I am passionate about teaching courses aligned with my research interests, particularly Deep Learning, Machine Learning, and Artificial Intelligence. Beyond advanced topics, I enjoy teaching foundational programming courses in C and C++ to first-year undergraduates, as well as Data Structures and Algorithms, which remains one of my favourite courses to teach. I have also taught Database Systems with hands-on MySQL labs and would be interested in teaching it again with an emphasis on relational learning approaches.

Looking ahead, I would like to develop and teach courses on Graph Representation Learning, Geometric Deep Learning, Neuro-Symbolic Learning, Theory of Large Language Models, Learning from Relational Databases, Mathematical and Computational Biology, Algorithms for Computational Biology, and AI for Healthcare and Medicine.

Academic and Research Services

I contribute to research indirectly, by reviewing for journals and conferences. Some of my activities are listed here.

Journal Editor

PLOS One (since Jan 2023)

Journal Reviewer

(Frequent ones are marked ‘*’, where I have reviewed at least 4 different full papers. I have now stopped reviewing for some of these journals due to: (a) quality of submissions, (b) too busy doing other things, (c) tired of incentive-less reviewing.)

IEEE Transactions on Industrial Electronics (TIE)*
IEEE Transactions on Cybernetics (CYB-IEEE)*
Neural Processing Letters (NEPL)*
IEEE Systems Journal (IEEE-SJ)
IEEE Transactions on Network Science and Engineering (TNSE)
IEEE Consumer Electronics Magazine (CEMAG)
IEEE Access
Scientific Data (Nature)
Information Fusion (INFFUS)*
Cognitive Systems Research (COGSYS)*
Applied Soft Computing (ASOC)
Frontiers in Neuroscience (Front. Neurosci.)
Soft Computing (SOCO)
Computers & Security (COSE)
Computer Methods and Programs in Biomedicine (CMPB)
Paladyn Journal of Behavioral Robotics (PJBR)
Robotica
Ain Shams Engineering Journal (ASEJ)
Proceedings of the National Academy of Sciences India (NASA)
International Journal Intelligent Robotics and Applications (JIRA)
International Journal of Image and Data Fusion (TIDF)
International Journal of Communication Systems (IJCS)
Data Technologies and Applications (DTA)
CAAI Transactions on Intelligence Technology (CIT)
Journal of Experimental & Theoretical Artificial Intelligence (TETA)

Conference Activities

Member of program committee, IJCAI: 2022–2024
Member of program committee, AI for Social Good Track, IJCAI: 2023–2024
Member of program committee, DASFAA: 2022
Member of program committee, AAAI-CLeaR: 2022
Member of technical committee, IJCNN: 2019–2023
Member of program committee, ICANN: 2019–2021
Member of program committee, ILP: 2019
Member of organizing committee, ICCI: 2017 (IIT Kanpur)

What else, if it matters to you?

I was born and raised in a remote village of Western Odisha, an Indian state rich in culture, vegetables and farming. My father (who is a Master in Arts and Bhagavad Gita) and extended paternal family are involved mostly in farming and small businesses. We speak a dialect of Odia language, known as Kosali (a folk language). My district is known for Odia folk music and the famous Sambalpuri Saree. I enjoy riding motorbikes (mostly fast-yet-careful racing). I play cricket; I am a batsman and a wicketkeeper. That view of the game from behind the stumps is extraordinary. I have occasionally played football; I am a goalkeeper. I love swimming in the sea (The Arabian sea is my neighbour) and known rivers. Trekking is something that always excites me. Watching movies on weekends has always been there since my undergrad studies (sometimes, I do binge-watch). In addition to all these fun stuff, I am also interested in philosophy, spirituality, farming, and poverty prevention.

Websites, I regularly visit: