arXiv subjects

Jithin George

Publications and source records attributed to Jithin George.

4 recordsLinked to original sources

arXiv2026

Cognis: Context-Aware Memory for Conversational AI Agents

LLM agents lack persistent memory, causing conversations to reset each session and preventing personalization over time. We present Lyzr Cognis, a unified memory architecture for conversational AI agents that addresses this limitation through a multi-stage retrieval pipeline. Cognis combines a dual-store backend pairing OpenSearch BM25 keyword matching with Matryoshka vector similarity search, fused via Reciprocal Rank Fusion. Its context-aware ingestion pipeline retrieves existing memories before extraction, enabling intelligent version tracking that preserves full memory history while keeping the store consistent. Temporal boosting enhances time-sensitive queries, and a BGE-2 cross-encoder reranker refines final result quality. We evaluate Cognis on two independent benchmarks -- LoCoMo and LongMemEval -- across eight answer generation models, demonstrating state-of-the-art performance on both. The system is open-source and deployed in production serving conversational AI applications.

cs.CL↗

arXiv2026

The Six Sigma Agent: Achieving Enterprise-Grade Reliability in LLM Systems Through Consensus-Driven Decomposed Execution

Large Language Models demonstrate remarkable capabilities yet remain fundamentally probabilistic, presenting critical reliability challenges for enterprise deployment. We introduce the Six Sigma Agent, a novel architecture that achieves enterprise-grade reliability through three synergistic components: (1) task decomposition into a dependency tree of atomic actions; (2) micro-agent sampling where each task is executed n times in parallel across diverse LLMs to generate independent outputs; and (3) consensus voting with dynamic scaling, clustering outputs and selecting the answer from the winning cluster with maximum votes. We prove that sampling n independent outputs with error rate p achieves system error O(p^{ceil(n/2)}), enabling exponential reliability gains. Even using cheaper models with 5% per-action error, consensus voting with 5 agents reduces error to 0.11%; dynamic scaling to 13 agents achieves 3.4 DPMO (Defects Per Million Opportunities), the Six Sigma standard. Evaluation across three enterprise use cases demonstrates a 14,700x reliability improvement over single-agent execution while reducing costs by 80%. Our work establishes that reliability in AI systems emerges from principled redundancy and consensus rather than model scaling alone.

cs.AI↗

arXiv2017

Effect of Transmission Impairments in CO-OFDM Based Elastic Optical Network Design

Coherent Optical Orthogonal Frequency Division Multiplexing (CO-OFDM) based Elastic Optical Network (EON) is one of the emerging technologies being considered for next generation high data rate optical network systems. Routing and Spectrum Allocation (RSA) is an important aspect of EON. Apart from spectral fragmentation created due to spectrum continuity and contiguity constraints of RSA, transmission impairments such as shot noise, amplified spontaneous emission (ASE) beat noise due to coherent detection, crosstalk in cross-connect (XC), nonlinear interference, and filter narrowing, limit the transmission reach of optical signals in EON. This paper focuses on the cross-layer joint optimization of delay-bandwidth product, fragmentation and link congestion for RSA in CO-OFDM EON while considering the effect of physical layer impairments. First, we formulate an optimal Integer Linear Programming (ILP) that achieves load-balancing in presence of transmission impairments and minimizes delay-bandwidth product along with fragmentation. We next propose a heuristic algorithm for large networks with two different demand ordering techniques. We show the benefits of our algorithm compared to the existing load balancing algorithm.

cs.NI↗

arXiv2011

The Outer Limits of Galaxy Clusters: Observations to the Virial Radius with Suzaku, XMM, and Chandra

The outskirts of galaxy clusters, near the virial radius, remain relatively unexplored territory and yet are vital to our understanding of cluster growth, structure, and mass. In this presentation, we show the first results from a program to constrain the state of the outer intracluster medium (ICM) in a large sample of galaxy clusters, exploiting the strengths of three complementary X-ray observatories: Suzaku (low, stable background), XMM-Newton (high sensitivity), and Chandra (good spatial resolution). By carefully combining observations from the cluster core to beyond r_200, we are able to identify and reduce systematic uncertainties that would impede our spatial and spectral analysis using a single telescope. Our sample comprises nine clusters at z ~ 0.1-0.2 fully covered in azimuth to beyond r_200, and our analysis indicates that the ICM is not in hydrostatic equilibrium in the cluster outskirts, where we see clear azimuthal variations in temperature and surface brightness. In one of the clusters, we are able to measure the diffuse X-ray emission well beyond r_200, and we find that the entropy profile and the gas fraction are consistent with expectations from theory and numerical simulations. These results stand in contrast to recent studies which point to gas clumping in the outskirts; the extent to which differences of cluster environment or instrumental effects factor in this difference remains unclear. From a broader perspective, this project will produce a sizeable fiducial data set for detailed comparison with high-resolution numerical simulations.

astro-ph.CO↗