arXiv subjects

Daniel Silva

Publications and source records attributed to Daniel Silva.

10 recordsLinked to original sources

arXiv2026

Structural Validation of LLM-Generated Microservice Decompositions Using Source-Code Dependencies

Decomposing monolithic systems into microservices is a key activity in software modernization. Although Large Language Models (LLMs) can generate semantically plausible decompositions from textual requirements, it remains unclear whether these proposals preserve the structural dependencies implemented in the source code. This paper evaluates the structural adherence of microservice decompositions generated by OpenAI o3 for the PetClinic and Bookstore systems. We propose an automated validation pipeline based on static dependency analysis and compare zero-shot and few-shot prompting using dependency preservation (TPD) and dependency violation (TVD) metrics. A robustness analysis was conducted to control for differences in class-to-service mapping coverage. After normalization, both prompting strategies produced equivalent structural adherence, achieving TPD values of 68.0% (PetClinic) and 83.3% (Bookstore). The findings demonstrate that structural evaluations of LLM-generated decompositions should explicitly control for mapping coverage, as apparent differences between prompting strategies may otherwise reflect methodological bias rather than genuine architectural quality.

cs.SE↗

arXiv2026

Building Customer Support AI Agents at 100M-User Scale: An Evaluation-Driven Framework

The rapid rise in LLM capabilities has made AI agents increasingly viable across a broad range of tasks. Among the most promising applications is building production-ready customer-facing agents, a challenge that demands coordinated excellence in evaluation methodology, context engineering, training, and online measurement. Yet these critical pillars are typically developed in isolation, creating blind spots that only surface after deployment. In this paper, we present a unified framework that bridges offline development with online impact for customer support AI agents at Nubank, a company with 100M+ users. Our approach integrates several key components: (1) structured context engineering tailored to customer support agents, (2) systematic human-in-the-loop prompt iteration, (3) rigorous LLM judge evaluation with measured inter-rater agreement and GEPA optimization for consistency, and (4) ideation-to-production validation. A central insight is that evaluation-pipeline quality directly determines iteration velocity. We present results from five production deployments spanning distinct domains: card delivery, debt management, credit-limit support, card management, and product explanation. These deployments deliver consistent customer-satisfaction gains while substantially accelerating iteration. In our card-delivery deployment, large-scale A/B testing yields a 37 percentage-point improvement in AI transactional Net Promoter Score and a 29 percentage-point gain in self-service rate over prior agent variants, alongside a strong correlation between offline simulation metrics and online outcomes, demonstrating that eval-driven development reliably predicts production impact. On most use cases, AI satisfaction reaches within a few percentage points of expert human agents.

cs.CL↗

arXiv2025

Effective Quantization of Muon Optimizer States

The Muon optimizer, based on matrix orthogonalization, has recently shown faster convergence and better computational efficiency over AdamW in LLM pre-training. However, the memory overhead of maintaining high-precision optimizer states remains a challenge for large-scale deployment. In this paper, we introduce the 8-bit Muon optimizer using blockwise quantization. In extensive Chinchilla-optimal experiments on pre-training models of up to 2.7B in size and fine-tuning them for instruction following, we demonstrate that 8-bit Muon achieves parity with Muon in terms of validation loss and downstream benchmarks, while achieving up to a 62\% reduction in optimizer state footprint. Crucially, we show that Muon's update mechanism is uniquely compatible with a simple linear quantization scheme, bypassing the complex dynamic scaling required for quantized AdamW. We supplement our empirical findings with a theoretical analysis of Muon's robustness to quantization noise.

cs.LG↗

arXiv2025

Your Spending Needs Attention: Modeling Financial Habits with Transformers

Predictive models play a crucial role in the financial industry, enabling risk prediction, fraud detection, and personalized recommendations, where slight changes in core model performance can result in billions of dollars in revenue or losses. While financial institutions have access to enormous amounts of user data (e.g., bank transactions, in-app events, and customer support logs), leveraging this data effectively remains challenging due to its complexity and scale. Thus, in many financial institutions, most production models follow traditional machine learning (ML) approaches by converting unstructured data into manually engineered tabular features. Conversely, other domains (e.g., natural language processing) have effectively utilized self-supervised learning (SSL) to learn rich representations from raw data, removing the need for manual feature extraction. In this paper, we investigate using transformer-based representation learning models for transaction data, hypothesizing that these models, trained on massive data, can provide a novel and powerful approach to understanding customer behavior. We propose a new method enabling the use of SSL with transaction data by adapting transformer-based models to handle both textual and structured attributes. Our approach, denoted nuFormer, includes an end-to-end fine-tuning method that integrates user embeddings with existing tabular features. Our experiments demonstrate improvements for large-scale recommendation problems at Nubank. Notably, these gains are achieved solely through enhanced representation learning rather than incorporating new data sources.

cs.IR↗

arXiv2024

HoneyGAN Pots: A Deep Learning Approach for Generating Honeypots

This paper investigates the feasibility and effectiveness of employing Generative Adversarial Networks (GANs) for the generation of decoy configurations in the field of cyber defense. The utilization of honeypots has been extensively studied in the past; however, selecting appropriate decoy configurations for a given cyber scenario (and subsequently retrieving/generating them) remain open challenges. Existing approaches often rely on maintaining lists of configurations or storing collections of pre-configured images, lacking adaptability and efficiency. In this pioneering study, we present a novel approach that leverages GANs' learning capabilities to tackle these challenges. To the best of our knowledge, no prior attempts have been made to utilize GANs specifically for generating decoy configurations. Our research aims to address this gap and provide cyber defenders with a powerful tool to bolster their network defenses.

cs.CR↗

arXiv2023

Topological Symmetry Groups of the Petersen graphs

The {\em topological symmetry group} of an embedding $\Gamma$ of an abstract graph $\gamma$ in $S^3$ is the group of automorphisms of $\gamma$ which can be realized by homeomorphisms of the pair $(S^3, \Gamma)$. These groups are motivated by questions about the symmetries of molecules in space. The Petersen family of graphs is an important family of graphs for many problems in low dimensional topology, so it is desirable to understand the possible groups of symmetries of their embeddings in space. In this paper, we find all the groups which can be realized as topological symmetry groups for each of the graphs in the Petersen Family. Along the way, we also complete the classification of the realizable topological symmetry groups for $K_{3,3}$.

math.GT↗

arXiv2023

Multicritical Bifurcation and First-order Phase Transitions in a Three-dimensional Blume-Capel Antiferromagnet

We present a detailed study by Monte Carlo simulations and finite-size scaling analysis of the phase diagram and ordered bulk phases for the three-dimensional Blume-Capel antiferromagnet in the space of temperature and magnetic and crystal fields (or two chemical potentials in an equivalent lattice-gas model with two particle species and vacancies). The phase diagram consists of surfaces of second- and first-order transitions that enclose a ``volume'' of ordered phases in the phase space. At relatively high temperatures, these surfaces join smoothly along a line of tricritical points, and at zero magnetic field we obtain good agreement with known values for tricritical exponent ratios [Y. Deng and H.W.J. Bl{\"o}te, Phys. Rev. E {70}, 0456111 (2004)]. In limited field regions at lower temperatures (symmetric under reversal of the magnetic field), the tricritical line for this three-dimensional model bifurcates into lines of critical endpoints and critical points, connected by a surface of weak first-order transitions {inside} the region of ordered phases. This phenomenon is not seen in the two-dimensional version of the same model. We confirm the location of the bifurcation as previously reported [Y.L. Wang and J.D. Kimel, J. Appl. Phys. {69}, 6176 (1991)], and we identify the phases separated by this first-order surface as antiferromagnetically (three-dimensional checker-board) ordered with different vacancy densities. We visualize the phases by real-space snapshots and by structure factors in the three-dimensional space of wave vectors.

cond-mat.stat-mech↗

arXiv2019

Sharing and Learning Alloy on the Web

We present Alloy4Fun, a web application that enables online editing and sharing of Alloy models and instances, to be used mainly in an educational context. By introducing the notion of secret paragraphs and commands in the models, it also allows the distribution and automatic evaluation of simple specification challenges, a useful mechanism that enables students to learn relational logic at their own pace. Alloy4Fun stores all versions of shared and analyzed models, as well as derivation trees that depict how those models evolved over time: this wealth of information can be mined by researchers or tutors to identify, for example, learning breakdowns in the class or typical mistakes made by students and other Alloy users. A beta version of Alloy4Fun was already used in two formal methods courses, and we present some results of this preliminary evaluation.

cs.CY↗

arXiv2019

Complete catalog of ground-state diagrams for the general three-state lattice-gas model with nearest-neighbor interactions on a square lattice

The ground states of the general three-state lattice-gas (equivalently, $S=1$ Ising) model with nearest-neighbor interactions on a square lattice are explored in the full, five-dimensional parameter space of three interaction constants and two generalized chemical potentials or fields. The resulting, complete catalog of fifteen topologically different ground-state diagrams (zero-temperature phase diagrams) is discussed in both lattice-gas and Ising-spin language. The results extend those of a recent study in a reduced parameter space [V. F. Fefelov, et al., Phys. Chem. Chem. Phys., 2018, 20, 10359--10368], which identified six different ground-state diagrams.

cond-mat.stat-mech↗

arXiv2013

Hierarchical Nystrom Methods for Constructing Markov State Models for Conformational Dynamics

Markov state models (MSMs) have become a popular approach for investigating the conformational dynamics of proteins and other biomolecules. MSMs are typically built from numerous molecular dynamics simulations by dividing the sampled configurations into a large number of microstates based on geometric criteria. The resulting microstate model can then be coarse-grained into a more understandable macro state model by lumping together rapidly mixing microstates into larger, metastable aggregates. However, finite sampling often results in the creation of many poorly sampled microstates. During coarse-graining, these states are mistakenly identified as being kinetically important because transitions to/from them appear to be slow. In this paper we propose a formalism based on an algebraic principle for matrix approximation, i.e. the Nystrom method, to deal with such poorly sampled microstates. Our scheme builds a hierarchy of microstates from high to low populations and progressively applies spectral clustering on sets of microstates within each level of the hierarchy. It helps spectral clustering identify metastable aggregates with highly populated microstates rather than being distracted by lowly populated states. We demonstrate the ability of this algorithm to discover the major metastable states on two model systems, the alanine dipeptide and TrpZip2.

q-bio.BM↗