Cascade: Exploiting SLO-Aware latency budget for fair and high goodput LLM inference serving

The reasoning and agentic capabilities of large language models have expanded the range of applications they support, from short interactive exchanges to long, compute-heavy requests. LLM serving platforms today define response-latency service-level objectives, even though requests within the same service can differ by orders of magnitude in input length, generation length, execution cost, and the availability of reusable KV-cache state. As a result, requests governed by the same service level objective have different urgency: after accounting for the time required to execute them, some have substantial latency headroom while others have almost none. We define this headroom---the difference between a request's service level objective and its predicted remaining service time---as its per-request latency budget. We present Cascade, an LLM serving system that estimates and continuously updates this budget from request characteristics, KV-cache state, and current system load. Unlike prior SLO-aware schedulers that use deadlines to govern request ordering alone, Cascade uses a single per-request budget to jointly coordinate request scheduling and KV-cache management across the memory hierarchy. Its scheduler prioritizes requests with little remaining budget, while its memory manager uses the same budget to decide whether non-resident KV state should be restored or prefetched from a deeper tier, retained in HBM, or recomputed. By directing queueing and data-movement overhead toward requests that can absorb it, Cascade improves SLO-satisfied goodput while preserving fairness across heterogeneous request classes. On production traces across three large language models, Cascade improves goodput by up to2.4x and reduces SLO violations by 40% relative to the default vLLM first-come, first-served scheduler.

Transported Memory Networks accelerating Computational Fluid Dynamics

In recent years, augmentation of differentiable PDE solvers with neural networks has shown promising results, particularly in fluid simulations. However, most approaches rely on convolutional neural networks and custom solvers operating on Cartesian grids with efficient access to cell data. This particular choice poses challenges for industrial-grade solvers that operate on unstructured meshes, where access is restricted to neighboring cells only. In this work, we address this limitation using a novel architecture, named Transported Memory Networks. The architecture draws inspiration from both traditional turbulence models and recurrent neural networks, and it is fully compatible with generic discretizations. Our results show that it is point-wise and statistically comparable to, or improves upon, previous methods in terms of both accuracy and computational efficiency.

FreSeg: Frenet-Frame-based Part Segmentation for 3D Curvilinear Structures

Part segmentation is a crucial task for 3D curvilinear structures like neuron dendrites and blood vessels, enabling the analysis of dendritic spines and aneurysms with scientific and clinical significance. However, their diversely winded morphology poses a generalization challenge to existing deep learning methods, which leads to labor-intensive manual correction. In this work, we propose FreSeg, a framework of part segmentation tasks for 3D curvilinear structures. With Frenet-Frame-based point cloud transformation, it enables the models to learn more generalizable features and have significant performance improvements on tasks involving elongated and curvy geometries. We evaluate FreSeg on 2 datasets: 1) DenSpineEM, an in-house dataset for dendritic spine segmentation, and 2) IntrA, a public 3D dataset for intracranial aneurysm segmentation. Further, we will release the DenSpineEM dataset, which includes roughly 6,000 spines from 69 dendrites from 3 public electron microscopy (EM) datasets, to foster the development of effective dendritic spine instance extraction methods and, consequently, large-scale connectivity analysis to better understand mammalian brains.

Detecting Synapse Location and Connectivity by Signed Proximity Estimation and Pruning with Deep Nets

Synaptic connectivity detection is a critical task for neural reconstruction from Electron Microscopy (EM) data. Most of the existing algorithms for synapse detection do not identify the cleft location and direction of connectivity simultaneously. The few methods that computes direction along with contact location have only been demonstrated to work on either dyadic (most common in vertebrate brain) or polyadic (found in fruit fly brain) synapses, but not on both types. In this paper, we present an algorithm to automatically predict the location as well as the direction of both dyadic and polyadic synapses. The proposed algorithm first generates candidate synaptic connections from voxelwise predictions of signed proximity generated by a 3D U-net. A second 3D CNN then prunes the set of candidates to produce the final detection of cleft and connectivity orientation. Experimental results demonstrate that the proposed method outperforms the existing methods for determining synapses in both rodent and fruit fly brain.

RhoanaNet Pipeline: Dense Automatic Neural Annotation

Reconstructing a synaptic wiring diagram, or connectome, from electron microscopy (EM) images of brain tissue currently requires many hours of manual annotation or proofreading (Kasthuri and Lichtman, 2010; Lichtman and Sanes, 2008; Seung, 2009). The desire to reconstruct ever larger and more complex networks has pushed the collection of ever larger EM datasets. A cubic millimeter of raw imaging data would take up 1 PB of storage and present an annotation project that would be impractical without relying heavily on automatic segmentation methods. The RhoanaNet image processing pipeline was developed to automatically segment large volumes of EM data and ease the burden of manual proofreading and annotation. Based on (Kaynig et al., 2015), we updated every stage of the software pipeline to provide better throughput performance and higher quality segmentation results. We used state of the art deep learning techniques to generate improved membrane probability maps, and Gala (Nunez-Iglesias et al., 2014) was used to agglomerate 2D segments into 3D objects. We applied the RhoanaNet pipeline to four densely annotated EM datasets, two from mouse cortex, one from cerebellum and one from mouse lateral geniculate nucleus (LGN). All training and test data is made available for benchmark comparisons. The best segmentation results obtained gave $V^\text{Info}_\text{F-score}$ scores of 0.9054 and 09182 for the cortex datasets, 0.9438 for LGN, and 0.9150 for Cerebellum. The RhoanaNet pipeline is open source software. All source code, training data, test data, and annotations for all four benchmark datasets are available at www.rhoana.org.