Hybrid Intent-Aware Personalization with Machine Learning and RAG-Enabled Large Language Models for Financial Services Marketing

Akhil Chandra Shanivendra

cs.LGMar 11, 2026

Hybrid Intent-Aware Personalization with Machine Learning and RAG-Enabled Large Language Models for Financial Services Marketing

Authors: Akhil Chandra Shanivendra

Organizations: Independent Researcher

Abstract

Personalized marketing in financial services requires models that can both predict customer behavior and generate compliant, context-appropriate content. This paper presents a hybrid architecture that integrates classical machine learning for segmentation, latent intent modeling, and personalization prediction with retrieval-augmented large language models for grounded content generation. A synthetic, reproducible dataset is constructed to reflect temporal customer behavior, product interactions, and marketing responses. The proposed framework incorporates temporal encoders, latent representations, and multi-task classification to estimate segment membership, customer intent, and product-channel recommendations. A retrieval-augmented generation layer then produces customer-facing messages constrained by retrieved domain documents. Experiments show that temporal modeling and intent features improve personalization accuracy, while citation-based retrieval reduces unsupported generation and supports auditability in regulated settings. The contribution is primarily architectural, demonstrating how predictive modeling and RAG-based generation can be combined into a transparent, explainable pipeline for financial services personalization.

Figures & tables

Figure 1: Architecture illustrating the fusion of static attributes and behavioral time series into the TL-ADGNPT Core Neural Network, feeding into the RAG Layer for content generation.

Category	Parameter / Setting	Value / Description
Dataset	Total Customers	35,000
	Observation Window	6 Months ( $K=6$ )
	Splits (Train/Val/Test)	70% / 15% / 15%
Features	Behavioral	Login, Transactions, Page Views, Delinquency
	Static	Age, Income, Tenure, Region
	Latent Variables	5 Segments, 5 Intent States

Table 1: Summary of Experimental Dataset and Model Configurations

Figure 2: Comparison of Normal Sequence (Control) vs. Shuffled Sequence (Experimental) to validate temporal dependency.

Model Setting	Overall (Avg)	Product	Channel	Timing	Intensity
TL-ADGNPT (Full Model)	0.9408	0.9349	0.8949	0.9561	0.9774
Ablation: No-Intent	0.9215	0.9180	0.8520	0.9450	0.9610
Ablation: No-Segment	0.9185	0.8750	0.9010	0.9320	0.9710
Baseline: No-Temporal	0.6370	0.6674	0.6090	0.5171	0.7545
Temporal Shuffled (Test)	0.6700	0.7146	0.5814	0.6943	0.6897

Table 2: TL-ADGNPT Predictive Performance (Macro-F1 Score)

Figure 3: Macro-F1 Score comparison between TL-ADGNPT (Full), Non-Temporal Baseline, and Temporal Shuffled.

Figure 4: Contribution of components across personalization dimensions.

Figure 5: Schematic representation of the Financial RAG Pipeline, illustrating the flow from inputs to the final cited explanation.

Metric	Result	Description
Response Rate	100%	Successfully generated a response
JSON Validity	99.2%	Output followed strict JSON schema
Citation Presence	98.5%	Message contained a citation
Citation Correctness	96.5%	Citations matched retrieved context
Avg. Message Length	$\sim$ 50 words	Concise, customer-facing length
Error Rate	0.8%	Minor formatting or citation failures

Table 3: RAG Generation Quality Metrics (N=400 Requests)

Financial Services

Financial services applies AI to banking, lending, payments, insurance, investment, and financial risk management.

See all 105 papers

From Clicks to Intent: Cross-Platform Session Embeddings with LLM-Distilled Taxonomy for Financial Services Recommendations
Jun 24, 2026Dianjing Fan, Yao Li, Kyaw Hpone Myint +4Sequential Recommendation Representation Learning
Capital One, McLean, VA, USA · Capital One, New York, NY, USA · Capital One, Boston, MA, USA
Benchmarking the Personalization Capabilities of Large Language Models
May 23, 2026Ashutosh Srivastava, Siddharth Yedlapati, Vinay Aggarwal +4Language Model Generation Evaluation Personalized Language Models
Adobe Media & Data Science Research
Automated Regulatory Compliance Question Answering in Financial Services with Domain-Adapted Retrieval-Augmented Generation
Sep 24, 2026Tobias Deußer, Abhishek Pillai, Aurelio F. Bariviera +5Retrieval-Augmented Generation Legal QA
University of Bonn, Bonn, Germany · Lamarr-Institute for Machine Learning and Artificial Intelligence, Bonn, Germany · Universitat Rovira i Virgili, Reus, Spain +1

Jun 24, 2026J/K move · Enter open · S save

From Clicks to Intent: Cross-Platform Session Embeddings with LLM-Distilled Taxonomy for Financial Services Recommendations

Dianjing Fan, Yao Li, Kyaw Hpone Myint, Dwipam Katariya+3

Capital One, McLean, VA, USA · Capital One, New York, NY, USA · Capital One, Boston, MA, USA

Sequential Recommendation Representation Learning LLM-Assisted Annotation Intent Classification Sequence Modeling

Open paper AskPDF arXiv

Sequential user behavior modeling is widely adopted in industrial recommender systems; however, significant gaps remain in financial services, where pre-login web interactions and authenticated in-app experiences differ drastically. Specifically, pre-login web users typically explore new products, whereas logged-in app users focus on account servicing. Due to the challenge of cross-channel entity resolution (e.g., matching anonymous web sessions to authenticated mobile accounts), web-based intent signals remain underutilized for post-authentication personalization. Existing methods for capturing web-based intent are often ad-hoc and narrow, lacking the flexibility to support both quantitative downstream recommendations and qualitative understanding at scale. In this work, we propose a scalable and dual-purpose intent prediction framework for web-based interactions and demonstrate its applicability for personalization. Our approach transforms raw web clickstreams into two outputs: a self-supervised Transformer encodes multi-modal clickstreams into a compact session embedding, while an LLM-based taxonomy generation and distillation pipeline produces interpretable intent labels. Our system demonstrates that self-supervised clickstream representations combined with LLM-distilled taxonomies can jointly serve quantitative tasks and qualitative understanding in production: on the mobile homepage tile ranking task, the session embedding improves macro Recall@1 by 1.88% and reduces Log Loss by 13.38% over production baselines. On the user conversion prediction task, the embedding outperforms the LLM labels by 4.3% on micro F1, while the distillation layer delivers interpretable labels at ultra-low latency with only a 7% performance drop.

Abstract

Financial Services

Explore similar work

From Clicks to Intent: Cross-Platform Session Embeddings with LLM-Distilled Taxonomy for Financial Services Recommendations

Benchmarking the Personalization Capabilities of Large Language Models

Automated Regulatory Compliance Question Answering in Financial Services with Domain-Adapted Retrieval-Augmented Generation