Banca de QUALIFICAÇÃO: REJANIO DAS NEVES DE MORAES FILHO
Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.STUDENT : REJANIO DAS NEVES DE MORAES FILHO
DATE: 25/02/2026
TIME: 09:00
LOCAL: Remoto
TITLE:
Multi-contextual data fusion with AI agents for vehicle recommendation
KEY WORDS:
Recommender Systems, Automotive Sector, Large Language Models (LLMs), AI Agents, Data Fusion.
PAGES: 45
BIG AREA: Engenharias
AREA: Engenharia Elétrica
SUMMARY:
The evolution of recommendation systems has been reshaping statistical approaches into cognitive models based on Large Language Models (LLMs), aiming to process the complexity of human reasoning. In the automotive sector, this migration is challenging because the decision-making process entails a high cognitive load and requires synthesizing heterogeneous variables, ranging from subjective preferences to technical safety indicators. The impasse lies in the trade-off between deep reasoning and operational efficiency, as the computational increase raises latency and costs without clear evidence of qualitative gains. Thus, this work evaluates the efficacy of an architecture based on Artificial Intelligence agents and LangGraph for personalized vehicle recommendation, grounded in the fusion of multi-contextual data. The methodology consisted of developing an agent that integrates three distinct databases (user profile, technical performance, and claims history). The evaluation was performed using local models (4B to 70B parameters) and commercial APIs, using ablation techniques to measure the impact of each component via latency and cost-per-token metrics. Partial results highlight that, although models with deep reasoning capabilities present a higher density of output tokens, there is a qualitative gain in recommendations when compared to conventional models.
COMMITTEE MEMBERS:
Presidente - 2885532 - IVANOVITCH MEDEIROS DANTAS DA SILVA
Interna - ***.640.764-** - MARIANNE BATISTA DINIZ DA SILVA - UFRN
Externo ao Programa - 2249146 - CARLOS MANUEL DIAS VIEGAS - UFRN