Deploying large language models (LLMs) often requires specialized hardware and complex frameworks, creating barriers for CPU-based environments with resource constraints. These systems, common in air-gapped or edge scenarios, lack support for maintenance due to security, budget, or technical limits. To address this, we introduce TranSQL⁺, a compiler that translates LLM inference into SQL queries, enabling deployment on relational databases. By converting transformer operations into relational algebra, TranSQL⁺ generates vector-oriented SQL queries that leverage native database features (buffer management, indexing) to manage computations without hardware accelerators or deep learning frameworks. Demonstrated with the LLaMA3.1 8B model on DuckDB, results show relational databases can effectively serve LLMs, reducing deployment barriers and expanding access to advanced AI.

Machine learning (ML) training data is often scattered across disparate collections of datasets, called <italic>data silos</italic>. This fragmentation poses a major challenge for data-intensive ML applications: integrating and transforming data residing in different sources demand a lot of manual work and computational resources. With data privacy constraints, data often cannot leave the premises of data silos; hence model training should proceed in a decentralized manner. In this work, we present a vision of bridging traditional data integration (DI) techniques with the requirements of modern machine learning systems. We explore the possibilities of utilizing metadata obtained from data integration processes for improving the effectiveness, efficiency, and privacy of ML models. Towards this direction, we analyze ML training and inference over data silos. Bringing data integration and machine learning together, we highlight new research opportunities from the aspects of systems, representations, factorized learning, and federated learning.

Machine learning (ML) practitioners and organizations are building model repositories of pre-trained models, referred to as model zoos. These model zoos contain metadata describing the properties of the ML models and datasets. The metadata serves crucial roles for reporting, auditing, ensuring reproducibility, and enhancing interpretability. Despite the growing adoption of descriptive formats like datasheets and model cards, the metadata available in existing model zoos remains notably limited. Moreover, existing formats have limited expressiveness, thus constraining the potential use of model repositories, extending their purpose beyond mere storage for pre-trained models. This paper proposes a unified metadata representation format for model zoos. We illustrate that comprehensive metadata enables a diverse range of applications, encompassing model search, reuse, comparison, and composition of ML models. We also detail the design and highlight the implementation of an advanced model zoo system built on top of our proposed metadata representation.

Machine learning (ML) training data is often scattered across disparate collections of datasets, called data silos. This fragmentation poses a major challenge for data-intensive ML applications: integrating and transforming data residing in different sources demand a lot of manual work and computational resources. With data privacy and security constraints, data often cannot leave the premises of data silos, hence model training should proceed in a decentralized manner. In this work, we present a vision of how to bridge the traditional data integration (DI) techniques with the requirements of modern machine learning. We explore the possibilities of utilizing metadata obtained from data integration processes for improving the effectiveness and efficiency of ML models. Towards this direction, we analyze two common use cases over data silos, feature augmentation and federated learning. Bringing data integration and machine learning together, we highlight new research opportunities from the aspects of systems, representations, factorized learning and federated learning.

Given a set of pre-trained Machine Learning (ML) models, can we solve complex analytic tasks that make use of those models by formulating ML inference queries? Can we mitigate different tradeoffs, e.g., high accuracy, low execution costs and memory footprint, when optimizing the queries? In this work we present different multi-objective ML inference query optimization strategies, and compare them on their usability, applicability, and complexity. We formulate Mixed-Integer-Programming-based (MIP) optimizers for ML inference queries that makes use of different objectives to find Pareto-optimal inference query plans.

Link prediction can be used to extract missing information, identify spurious interactions as well as forecast network evolution. Network embedding is a methodology to assign coordinates to nodes in a low-dimensional vector space. By embedding nodes into vectors, the link prediction problem can be converted into a similarity comparison task. Nodes with similar embedding vectors are more likely to be connected. Classic network embedding algorithms are random-walk-based. They sample trajectory paths via random walks and generate node pairs from the trajectory paths. The node pair set is further used as the input for a Skip-Gram model, a representative language model that embeds nodes (which are regarded as words) into vectors. In the present study, we propose to replace random walk processes by a spreading process, namely the susceptible-infected (SI) model, to sample paths. Specifically, we propose two susceptible-infected-spreading-based algorithms, i.e., Susceptible-Infected Network Embedding (SINE) on static networks and Temporal Susceptible-Infected Network Embedding (TSINE) on temporal networks. The performance of our algorithms is evaluated by the missing link prediction task in comparison with state-of-the-art static and temporal network embedding algorithms. Results show that SINE and TSINE outperform the baselines across all six empirical datasets. We further find that the performance of SINE is mostly better than TSINE, suggesting that temporal information does not necessarily improve the embedding for missing link prediction. Moreover, we study the effect of the sampling size, quantified as the total length of the trajectory paths, on the performance of the embedding algorithms. The better performance of SINE and TSINE requires a smaller sampling size in comparison with the baseline algorithms. Hence, SI-spreading-based embedding tends to be more applicable to large-scale networks.

Cameras are ubiquitous nowadays and video analytic systems have been widely used in surveillance, traffic control, business intelligence and autonomous driving. Some applications, e.g., detecting road congestion in traffic monitoring, require continuous and timely reporting of complex patterns. However, conventional complex event processing (CEP) systems fail to support video processing, while the existing video query languages offer limited support for expressing advanced CEP queries, such as iteration, and window. In this PhD research, we aim to develop systems and methods to alleviate these issues. In this paper, we first identify the need for an expressive CEP language which allows users to define queries over video streams, and receive fast, accurate results. To evaluate CEP queries on videos in real-time and with high accuracy, we explain how a streaming query engine can be designed to provide native support of machine learning (ML) models for fast and accurate inference on video streams. In addition, we describe a set of optimization problems that arise when ML models, with trade-offs in speed, accuracy, and cost, are part of a query plan. Finally, we describe how query plans on real-time videos can be optimized and deployed on edge devices with limited computational and network capabilities.