arXiv subjects

Sumit Saha

Publications and source records attributed to Sumit Saha.

6 recordsLinked to original sources

arXiv2026

KATANA: A Fast, Low-Power Mapping of Kalman Filters onto Edge NPUs for Real-Time Tracking

State estimation is the closed-loop core of every real-time tracking system, from radar surveillance and counter-UAV defense to autonomous driving and robotics. These deployments run on edge platforms, where defense systems mount on vehicles and drones, and civilian pipelines live on cars and handheld devices. Here, every additional watt of compute erodes mission duration or operational range. Two hard constraints follow: each new measurement must be fused before the next control cycle, and the total compute must fit within a strict battery and thermal power envelope. The Linear and Extended Kalman Filters (LKF, EKF) are dominant estimators on these systems, but today they execute almost exclusively on CPUs, which serialize multi-object tracking (MOT) updates, or on custom FPGA/ASIC accelerators that lengthen design cycles. Contemporary AI-PC SoCs, like the Intel Core Ultra Series 1 and 2, integrate a low-power, data-parallel Neural Processing Unit (NPU). We therefore ask whether the Kalman filter can be mapped onto this existing matrix engine to meet real-time and low-power budgets simultaneously, avoiding a dedicated accelerator and keeping the CPU and GPU free for primary workloads. We present KATANA, an NPU-aware optimization framework delivering the first end-to-end mapping of the LKF and EKF onto a commercial NPU, alongside a cross-platform characterization on shipping AI-PC silicon. KATANA applies three algebraic graph rewrites: subtract-to-add reformulation via a precomputed negative-projection matrix H_neg, static-shape tensor fusion, and block-diagonal batched parallelization, ensuring 100% of operations execute on the DPU matrix engine. On the Series 2, the optimized batched EKF reaches 223.35 FPS at 13.43 W active power, and the LKF reaches 408.73 FPS at 14.05 W, delivering up to a 97.9% reduction in dynamic energy versus the CPU implementation.

cs.AR↗

arXiv2026

Push Down Optimization for Distributed Multi Cloud Data Integration

Enterprises increasingly adopt multi cloud architectures to take advantage of diverse database engines, regional availability, and cost models. In these environments, ETL pipelines must process large, distributed datasets while minimizing latency and transfer cost. Push down optimization, which executes transformation logic within database engines rather than within the ETL tool, has proven highly effective in single cloud systems. However, when applied across multiple clouds, it faces challenges related to data movement, heterogeneous SQL engines, orchestration complexity, and fragmented security controls. This paper examines the feasibility of push down optimization in multi cloud ETL pipelines and analyzes its benefits and limitations. It evaluates localized push down, hybrid models, and data federation techniques that reduce cross cloud traffic while improving performance. A case study across Redshift and BigQuery demonstrates measurable gains, including lower end to end runtime, reduced transfer volume, and improved cost efficiency. The study highlights practical strategies that organizations can adopt to improve ETL scalability and reliability in distributed cloud environments.

cs.DC↗

arXiv2026

Scalable Sequential Recommendation under Latency and Memory Constraints

Sequential recommender systems must model long-range user behavior while operating under strict memory and latency constraints. Transformer-based approaches achieve strong accuracy but suffer from quadratic attention complexity, forcing aggressive truncation of user histories and limiting their practicality for long-horizon modeling. This paper presents HoloMambaRec, a lightweight sequential recommendation architecture that combines holographic reduced representations for attribute-aware embedding with a selective state space encoder for linear-time sequence processing. Item and attribute information are bound using circular convolution, preserving embedding dimensionality while encoding structured metadata. A shallow selective state space backbone, inspired by recent Mamba-style models, enables efficient training and constant-time recurrent inference. Experiments on Amazon Beauty and MovieLens-1M under a 10-epoch budget show that HoloMambaRec surpasses SASRec on both datasets, attains state-of-the-art ranking on MovieLens-1M, and trails only GRU4Rec on Amazon Beauty, all while maintaining substantially lower memory complexity. The design further incorporates forward-compatible mechanisms for temporal bundling and inference-time compression, positioning HoloMambaRec as a practical and extensible alternative for scalable, metadata-aware sequential recommendation.

cs.IR↗

arXiv2025

AI-Driven Cloud Resource Optimization for Multi-Cluster Environments

Modern cloud-native systems increasingly rely on multi-cluster deployments to support scalability, resilience, and geographic distribution. However, existing resource management approaches remain largely reactive and cluster-centric, limiting their ability to optimize system-wide behavior under dynamic workloads. These limitations result in inefficient resource utilization, delayed adaptation, and increased operational overhead across distributed environments. This paper presents an AI-driven framework for adaptive resource optimization in multi-cluster cloud systems. The proposed approach integrates predictive learning, policy-aware decision-making, and continuous feedback to enable proactive and coordinated resource management across clusters. By analyzing cross-cluster telemetry and historical execution patterns, the framework dynamically adjusts resource allocation to balance performance, cost, and reliability objectives. A prototype implementation demonstrates improved resource efficiency, faster stabilization during workload fluctuations, and reduced performance variability compared to conventional reactive approaches. The results highlight the effectiveness of intelligent, self-adaptive infrastructure management as a key enabler for scalable and resilient cloud platforms.

cs.DC↗

arXiv2025

An SLO Driven and Cost-Aware Autoscaling Framework for Kubernetes

Kubernetes provides native autoscaling mechanisms, including the Horizontal Pod Autoscaler, Vertical Pod Autoscaler, and node-level autoscalers, to enable elastic resource management for cloud-native applications. However, production environments frequently experience Service Level Objective violations and cost inefficiencies due to reactive scaling behavior, limited use of application-level signals, and opaque control logic. This paper investigates how Kubernetes autoscaling can be enhanced using AIOps principles to jointly satisfy SLO and cost constraints under diverse workload patterns without compromising safety or operational transparency. We present a gap-driven analysis of existing autoscaling approaches and propose a safe and explainable multi-signal autoscaling framework that integrates SLO-aware and cost-conscious control with lightweight demand forecasting. Experimental evaluation using representative microservice and event-driven workloads shows that the proposed approach reduces SLO violation duration by up to 31 percent, improves scaling response time by 24 percent, and lowers infrastructure cost by 18 percent compared to default and tuned Kubernetes autoscaling baselines, while maintaining stable and auditable control behavior. These results demonstrate that AIOps-driven, SLO-first autoscaling can significantly improve the reliability, efficiency, and operational trustworthiness of Kubernetes-based cloud platforms.

cs.SE↗

arXiv2022

Ferrocene as an iconic redox marker: from solution chemistry to molecular electronic devices

Ferrocene, since its discovery in 1951, has been extensively exploited as a redox probe in a variety of processes ranging from solution chemistry, medicinal chemistry, supramolecular chemistry, surface chemistry to solid-state molecular electronic and spintronic circuit elements to unravel electrochemical charge-transfer dynamics. Ferrocene represents an extremely chemically and thermally stable, and highly reproducible redox probe that undergoes reversible one-electron oxidation and reduction occurring at the interfaces of electrode/ferrocene solution in response to applied anodic and cathodic potentials, respectively. It has been almost 70 years after its discovery and has become one of the most widely studied and model organometallic compounds not only for probing electrochemical charge-transfer process but also as molecular building blocks for the synthesis of chiral organometallic catalysts, potential drug candidates, polymeric compounds, electrochemical sensors, to name a few. Ferrocene and its derivatives have been a breakthrough in many aspects due to its versatile reactivity, fascinating chemical structures, unconventional metal-ligand coordination, and the magic number of electrons (18 e-). The present review discusses the recent progress made towards ferrocene-containing molecular systems exploited for redox reactions, surface attachment, spin-dependent electrochemical process to probe spin polarization, photo-electrochemistry, and integration into prototype molecular electronic devices. Overall, the present reviews demonstrate a piece of collective information about the recent advancements made towards the ferrocene and its derivatives that have been utilized as iconic redox markers.

physics.chem-ph↗