An AI-Driven Smartphone Recommendation System Using RAG, LLMs, and Multisource Price Tracking
International Journal of Advanced Research in Science Communication and Technology · 29.06.2026 · 43
DOI: 10.48175/ijarsct-37307 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Advanced Research in Science Communication and Technology, online erschienen am 29.06.2026. Verfasst von Mrs. Abha Pathak und 5 weiteren Autor:innen. Die Arbeit nennt 39 Literaturangaben.
Abstract (Originalsprache)
Objective
Selecting a suitable smartphone has become challenging due to rapidly evolving specifications, market fragmentation, and limited user expertise. This paper presents the design and implementation of Transistor, an AI-driven smartphone recommendation system tailored for the Indian market.
Method
The system employs a hybrid pipeline combining Retrieval-Augmented Generation (RAG) with K-Nearest Neighbors (KNN) similarity search over dense product embeddings, orchestrated via a FastAPI backend. A structured, multi-schema PostgreSQL database stores 627 phone records across 17 brands, populated through an automated Firecrawl-based scraping pipeline and validated by a Human-in-the-Loop (HITL) review layer. A slot-filling conversational funnel collects up to 17 buyer-preference signals; a two-peak India recency model, a band-aware 5G connectivity scorer, and a graduated constraint-relaxation engine combine with semantic embedding retrieval to rank candidates.
Results
System demonstration on representative query scenarios covering budget-constrained, use-case-specific, spec-targeted, and comparative intents confirms that the pipeline produces contextually relevant, factually grounded recommendations with an average end-to-end response time of 3–5 seconds. The HITL validation layer is shown qualitatively to eliminate category-level factual errors present in raw-scraped data. Quantitative ranking evaluation against human-annotated relevance judgments is planned as future work. Contribution: The principal novelty is the integration of India-specific ranking signals—carrier 5G band compatibility (Jio, Airtel, Vi, BSNL), a dual-peak recency model reflecting India's two annual sales seasons, and a sale-season relevance multiplier—within a continuously refreshed, human-verified knowledge base of 627 phones across 17 brands. The system combines these India-market signals with semantic embedding retrieval and a slot-filling conversational funnel in a single deployable pipeline.
Zitieren
APA
Pathak, M. A., Mali, M. T., Rathod, S., Rane, N., Pimpare, S., & Rakh, A. (2026). An AI-Driven Smartphone Recommendation System Using RAG, LLMs, and Multisource Price Tracking. International Journal of Advanced Research in Science Communication and Technology, 43. https://doi.org/10.48175/ijarsct-37307
Vancouver
Pathak MA, Mali MT, Rathod S, Rane N, Pimpare S, Rakh A. An AI-Driven Smartphone Recommendation System Using RAG, LLMs, and Multisource Price Tracking. International Journal of Advanced Research in Science Communication and Technology. 2026:43. doi:10.48175/ijarsct-37307
BibTeX
@article{Pathak2026,
author = {Pathak, Mrs. Abha and Mali, Mrs. Tejaswini and Rathod, Sanket and Rane, Niraj and Pimpare, Swapnali and Rakh, Aditi},
title = {An AI-Driven Smartphone Recommendation System Using RAG, LLMs, and Multisource Price Tracking},
journal = {International Journal of Advanced Research in Science Communication and Technology},
year = {2026},
pages = {43},
doi = {10.48175/ijarsct-37307}
}
EUVIMED