IT Track

International Speaker:

Dr. Salil Kanhere
University of New South

National Speaker:

Dr. Pawan Goyal
Indian Institute of Technology Kharagpur, India

DC Track

International Speaker:

Dr. Sandeep Kulkarni
Michigan State University, USA

National Speaker:

Dr. Yogesh Simmahan
Indian Institute of Science, India

AI4Sc Track

International Speaker:

Swati Biswas
University of Texas at Dallas, USA

National Speaker:

U Deva Priyakumar
International Institute of Information Technology Hyderabad, India

DC TRACK

International Speaker:

Dr. Sandeep Kulkarni
Michigan State University, USA

Title: Asynchronous Algorithms

 

The talk will focus on the design and analysis of asynchronous algorithms, a class of concurrent and distributed computations that operate under weak memory models. In such settings, individual nodes or processes may observe outdated values of variables that other nodes or processes have modified. The objectives are twofold: to identify the necessary and sufficient conditions for correctness in this framework, and to develop algorithms for fundamental problems such as dominating set, vertex cover, and robot gathering. A notable feature of many of these algorithms is that they possess a self-stabilizing property, guaranteeing convergence to a desired target state even when the system starts from an arbitrary configuration

Dr. Sandeep S. Kulkarni is a Professor in the Department of Computer Science and Engineering at Michigan State University, USA, whose research focuses on distributed systems, fault tolerance, self-stabilization, and formal methods. He is widely known for his contributions to the design of correct-by-construction and self-recovering distributed systems, with an emphasis on rigorous reasoning and automated synthesis of fault-tolerant programs. His work bridges theory and practice, enabling systems to recover from transient faults without external intervention. Dr. Kulkarni has published extensively in leading journals and conferences, served on numerous program committees, and is deeply involved in mentoring students and advancing reliable and resilient computing methodologies .

National Speaker:

Dr. Yogesh Simmahan
Indian Institute of Science, India

Title: Distributed Intelligence on Dynamic Graphs: Partitioning, Mining and Learning at Scale

Modern applications spanning fintech to social and transportation networks require low latency analytics and learning over massive, continuously evolving graphs. This talk offers a systems-focused approach for distributed and streaming graph intelligence that addresses partitioning, temporal mining, and GNN training/inference. We will first present TriParts, a community aware streaming edge partitioner that maximizes local triangles to preserve structure within partitions, scaling to over 500M updates per second and boosting locality for downstream analytics. Next, we will discuss TARIS, a distributed runtime for time respecting algorithms on temporal/streaming graphs, which formalizes incremental computation over update windows to match batch semantics and enables scalable temporal mining on high rate streams. Building on these innovations, Ripple enables incremental GNN inference on streaming graphs via delta-propagation of linear aggregation, achieving up to 28k updates/sec on a single server by avoiding redundant neighborhood recomputation. Finally, I will cover OpES, an optimized federated GNN training framework that uses remote embedding pruning and communication-compute overlap to reduce memory/communication footprints and shorten time to accuracy, without sacrificing model quality. Together, these advances chart a path to scalable intelligence on dynamic graphs, making real time graph mining, learning, and serving feasible at web scale for practical applications.

Yogesh Simmhan is an Associate Professor in the Department of Computational and Data Sciences at the Indian Institute of Science, Bangalore. His research explores scalable software platforms, algorithms, and applications on distributed systems. These span Cloud and Edge Computing, Temporal Graph Processing, and Scalable Machine Learning to support emerging Big Data and Internet of Things (IoT) applications. He has published over 100 peer-reviewed papers, won the Best Paper Award at the IEEE International Conference on Cloud Computing (CLOUD) 2019, the IEEE TCSC SCALE Challenge Award in 2019 and 2012, the Distinguished Paper award at EuroPar 2018, and the IEEE/ACM Supercomputing HPC Storage Challenge Award in 2008. He is the recipient of the Swarna Jayanti Fellowship in 2019 and the IEEE TCSC Award for Excellence in Scalable Computing (Mid-career Researcher) in 2020. He is an Associate Editor of Future Generation Computing Systems (FGCS) and earlier served as Associate Editor-in-Chief of the Journal of Parallel and Distributed Systems (JPDC), as an Associate Editor of the IEEE Transactions on Cloud Computing, and as a member of the IEEE Future Directions Initiative on Big Data. Yogesh has a Ph.D. in Computer Science from Indiana University, Bloomington, and was previously a Research Assistant Professor at the University of Southern California (USC), Los Angeles, and a Postdoc at Microsoft Research, San Francisco. He is a Distinguished Member of ACM, a Distinguished Contributor of the IEEE Computer Society, and serves on the ACM India Executive Council.

IT TRACK

International Speaker:

Dr. Salil Kanhere
University of New South

Title: Cracking the Code: Detecting the Next Generation of Social Bots

Social bots have evolved into highly sophisticated actors within online ecosystems, shaping public discourse, amplifying misinformation, and executing coordinated campaigns at unprecedented scale. This talk presents two complementary investigations into their behaviour and detection. First, we introduce BotSSCL, a self-supervised contrastive learning framework that detects human-like bots by leveraging social network-inspired objectives to enhance representation separability. BotSSCL achieves robust, generalizable embeddings, outperforming state-of-the-art supervised and unsupervised baselines by up to 8% in F1 score. Second, we examine TBTrackerX, a measurement-driven approach to uncover trigger-based (TB) bot campaigns—automated agents that activate in response to specific keywords to engage in deceptive financial or illicit activities. Our large-scale analysis reveals a coordinated TB ecosystem characterised by context-dependent responses, intermittent activity, and strategic dormancy to evade detection. Together, these studies provide a comprehensive view of advanced bot operations and highlight the urgent need for adaptive, resilient, and trustworthy detection strategies in dynamic social media landscapes.

Salil Kanhere is a Professor at the School of Computer Science and Engineering at UNSW Sydney, Australia. His research interests include cybersecurity, mobile computing, IoT, blockchain, and applied machine learning. He has published over 400 peer-reviewed articles and leads several government and industry-funded research projects in these fields. He is an IEEE Fellow and ACM Distinguished Member. He received the Friedrich Wilhelm Bessel Research Award (2020) and the Humboldt Research Fellowship (2014) from the Alexander von Humboldt Foundation in Germany, along with 12 Best Paper Awards. He serves on the advisory board of three SMEs and has held visiting positions at RWTH Aachen, I2R Singapore, Technical University Darmstadt, the University of Zurich, and Graz University of Technology. Salil is the Editor in Chief of the Ad Hoc Networks journal and an Associate Editor of IEEE Transactions on Network and Service Management, Computer Communications, and Pervasive and Mobile Computing. He has participated in organising committees for several IEEE/ACM international conferences and is a steering committee member for IEEE LCN and IEEE ICBC. Salil has also co-authored two books.

National Speaker:

Dr. Pawan Goyal
Indian Institute of Technology Kharagpur, India

Title: Sanskrit and Computational Linguistics: Analysis and Generation of Sanskrit verses

The talk will focus on how Sanskrit manuscripts can be made more accessible to end users through natural language technologies. The morphological richness, compounding, free word orderliness, and low-resource nature of Sanskrit are recognised as significant challenges for the development of deep learning solutions. Fundamental tasks crucial to the development of robust NLP technology for Sanskrit are identified, including word segmentation, morphological parsing, dependency parsing, and syntactic linearisation, as well as challenging problems such as Sanskrit poetry generation. A brief presentation will be given of a framework using Energy-Based Models for multiple structured prediction tasks in Sanskrit. In this framework, a graph is expected as input; relevant linguistic information is encoded in the nodes, and the edges indicate the associations between them. Typically, state-of-the-art models for morphosyntactic tasks in morphologically rich languages still rely on hand-crafted features for performance, whereas here the learning of the feature function is automated. The feature function that is learnt, together with the search space that is constructed, is used to encode the relevant linguistic information for the tasks considered. By this means, the training data requirements are reduced substantially, to as low as 10% of the data required by neural state-of-the-art models.

Next, recent works that leverage the latest advances in deep learning for Sanskrit NLP will be discussed, along with interesting future directions in Sanskrit Computational Linguistics. One interesting work to be discussed addresses the question: Can LLMs be adapted for structured poetic generation in a low-resource, morphologically rich language such as Sanskrit? Current progress will be described, including the introduction of a dataset designed for the translation of English prose into structured Sanskrit verse, with strict adherence to classical metrical patterns, particularly the Anuṣṭubh meter. A range of generative models—both open-source and proprietary—are evaluated under multiple settings. Specifically, constrained decoding strategies and instruction-based fine-tuning tailored to metrical and semantic fidelity are explored. Using the proposed decoding approach, over 99% accuracy is achieved in producing syntactically valid poetic forms, and general-purpose models are substantially outperformed in terms of meter conformity. Meanwhile, instruction-tuned variants are shown to exhibit improved alignment with source meaning and poetic style, as indicated by human assessments, albeit at the cost of marginal trade-offs in metrical precision

Dr. Pawan Goyal is a Professor in the Department of Computer Science and Engineering at the Indian Institute of Technology (IIT) Kharagpur, India. He is a prominent researcher in the fields of Natural Language Processing (NLP), Text Mining, and Sanskrit Computational Linguistics. A prolific contributor to his field, Dr. Goyal has published over 250 research papers in major international conferences and journals. He is a member of the Complex Network Research Group (CNeRG) and has served as a Senior Area Chair and Program Committee member for prestigious global forums such as EMNLP, AAAI, and AACL.

Dr. Goyal has been the recipient of several notable honors, including the 2020 INAE Young Engineer Award, the Google India AI/ML Research Award 2020, and the Facebook AI and Ethics Research Award 2019. He was also honored with the Faculty Excellence Award from IIT Kharagpur in 2022 and the ACM HyperText Ted Nelson Newcomer Award in 2021. Prior to joining IIT Kharagpur in 2013, he served as a post-doctoral fellow at INRIA Paris-Rocquencourt, where he contributed to the Sanskrit Heritage Site.

Throughout his career, Dr. Goyal has led numerous high-impact sponsored research projects funded by agencies such as MeitY, Microsoft India, and the I-Hub Foundation for Cobotics, focusing on areas like Large Language Models (LLMs) for legal assistance, multilingual dialogue systems, and goal-oriented healthcare interactions. He is also a dedicated educator, having developed popular NPTEL courses on Natural Language Processing and Deep Learning, and has supervised a wide range of doctoral and master’s theses. His academic journey includes a B.Tech. in Electrical Engineering from IIT Kanpur and a Ph.D. from the University of Ulster, UK, where his research focused on analytic knowledge discovery for information retrieval and text summarization.

AI4SC TRACK

International Speaker:

Dr. Swati Biswas
University of Texas at Dallas, USA

Title: Absolute Risk Prediction for Substance Use Disorders using Bayesian Machine Learning

Substance use disorders (SUDs) are a growing public health concern worldwide. Early substance use during adolescence is a strong precursor of SUDs in adulthood. Identifying individuals at high-risk before the onset of problematic use provides a critical window of opportunity for timely prevention and intervention strategies. A potentially valuable tool for this purpose is a risk prediction model that can take an individual’s profile and provide their personalized probability, i.e., absolute risk of developing an SUD within a user-specified time interval. Such absolute risk prediction models are widely used in clinical settings for various diseases and disorders yet are missing for SUDs. To fill this gap, we first develop an absolute risk prediction model for cannabis use disorder (CUD) using Bayesian machine learning. Next, we extend it to propose a joint absolute risk prediction model for CUD and alcohol use disorder (AUD) that accounts for the correlation between the two outcomes using a shared frailty. These models are trained on data from the National Longitudinal Study of Adolescent to Adult Health (Add Health) collected in the US. Key predictors in our models include demographic characteristics, delinquent behavior, peer influence, and personality traits such as conscientiousness and neuroticism. The models are validated on an independent dataset from Add Health as well as on an external dataset from the Christchurch Health and Development Study conducted in New Zealand. Both cross-validation and independent validation demonstrate good discrimination and calibration performances of the proposed models. By providing personalized and time-specific risks for adolescents and young adults who use alcohol and/or cannabis, these models can support clinical decision-making by identification of high-risk individuals enabling timely, targeted prevention strategies. Moreover, the proposed Bayesian learning methodologies offer novel ways for absolute risk prediction of other diseases and disorders.

Dr. Swati Biswas is a Professor of Statistics and Associate Head of the Department of Mathematical Sciences at the University of Texas (UT) at Dallas. She received her PhD from the Ohio State University and did her postdoctoral training at the UT MD Anderson Cancer Center. Her research interests are in biostatistics, in particular, statistical genetics, genetic epidemiology, cancer genetics, and risk prediction models. She has been PI on several grants from the National Institutes of Health for developing statistical methods for detecting gene-environment interactions, risk prediction for breast cancer, and meta-analysis of cancer risk conferred by genetic mutations. Several of her proposed models are currently in clinical use. More recently, she has been also developing statistical and machine learning models for risk prediction of substance use disorders. She is one of the founding members of the Texas Artificial Intelligence Research Institute at UT Dallas. She is recipient of Young Researcher Award from the International Indian Statistical Association and is an Elected Member of the International Statistical Institute.

National Speaker:

Dr. U Deva Priyakumar
Center for Computational Natural Sciences and Bio-informatics, IIIT Hyderabad, India

Title: Modern Artificial Intelligence for Drug Discovery

In recent years, modern AI/ML methods have revolutionized and have had phenomenal success in many technology areas such as computer vision, natural language processing, machine translation, speech recognition, autonomous driving, etc. Availability of high performance GPU accelerators, historical data and development of new algorithms/libraries have fueled such a surge. This has inspired chemists and biologists to apply these algorithms to problems in molecular science that are traditionally tackled using physics-based methods. The past few years have witnessed ML based solutions in molecular science that are more efficient and accurate compared to traditional methods. In this talk, I will initially introduce the philosophy of machine learning methods in the context of the three paradigms of research (experiments, theory and computations). I will also introduce the direct and inverse problems in the context of molecular/material design. Secondly, we will discuss development of ML algorithms for some direct problems that include predicting DFT energies/forces, solvation free energies and protein binding sites. In the last part of the talk, we will discuss inverse problem solving and explore some methods that may enable self-driving chemistry laboratories. We will discuss methods that are capable of generating novel molecules based on a chemist’s requirement (like ChatGPT for text), and how AI is capable of elucidating structures given molecular spectra as inputs. Finally, we will briefly discuss how the informatics based approaches could impact the way we do research in computational molecular science.

Deva finished his PhD from Pondicherry University/IICT Hyderabad followed by a postdoctoral fellowship at the University of Maryland. He is currently Professor and Dean (R&D) at IIIT Hyderabad. He was the founding Project Director of IHub-Data, a Technology Innovation Hub on Data Driven Technologies. His major research interests are in the areas of applying computational methods for studying chemical and biological systems/processes. Recently, his group has made significant contributions in applying modern AI/ML techniques for molecular science research. He has received Indian National Science Academy Young Scientist Medal, Young Associate Fellowship from Indian Academy of Sciences, Innovative Young Biotechnologist Award, Distinguished Lectureship Award by CSJ, JSPS Fellowship, Chemical Research Society of India Medal and Google Impact Scholar award.