Halchal E-Commerce Platform with CI/CD Pipeline and Automated Pricing System

Halchal E-Commerce Platform with CI/CD Pipeline and Automated Pricing System

Hrishikesh Nagargoje, Devesh Surana, Shreyash Hubale, Anas Shaikh, Prof. Neelam Jain

Artificial Intelligence & Data Science Department Ajeenkya D.Y. Patil School of Engineering, Lohegaon, Pune – 412105, Maharashtra, INDIA

Abstract – Recent advancements in artificial intelligence have transformed digital commerce systems, enabling intelligent automation for intelligent systems that improve pricing strategies, inventory planning, and customer interaction for manufacturing businesses. This paper proposes an AI-enabled e-commerce platform developed for Halchal Industries to digitize traditional sales operations through intelligent automation and controlled decision support. The proposed system integrates an AI-assisted pricing mechanism that generates price recommendations based on purchase quantity, customer region, seasonal demand patterns, and competitor pricing references while maintaining human-in-the-loop administrative approval to ensure transparency and business control. A machine learningbased demand forecasting module using Random Forest regression analyzes historical sales data to predict seasonal demand trends and support inventory planning. In addition, a chatbot powered by a pretrained natural language processing model enhances customer engagement by providing product guidance and navigation assistance. The platform is implemented using a full-stack architecture with React.js, Node.js, Express.js, and MongoDB, along with CI/CD automation to improve deployment reliability and development efficiency. The proposed framework demonstrates a scalable and practical approach to integrating artificial intelligence with controlled business workflows in modern e-commerce systems.

Keywords: AI-enabled e-commerce, AI-assisted pricing, demand forecasting, inventory management, seasonal demand analysis, chatbot system, CI/CD automation, full-stack web application, MongoDB, Node.js.

1. INTRODUCTON

   The rapid growth of digital commerce has increased the need for intelligent and automated systems that can improve operational efficiency, pricing accuracy, and customer engagement, particularly for small and medium-scale manufacturing businesses. Traditional sales systems in such organizations often rely on manual processes for order handling, pricing decisions, and inventory planning, which limits scalability and responsiveness to changing market conditions. In industries such as agricultural manufacturing, where demand varies significantly across regions and seasons, static pricing and manual inventory management can lead to revenue loss, overstocking, or missed sales opportunities.

   The Halchal E-Commerce Platform is proposed as an AI-enabled digital solution that bridges this gap by combining e-commerce automation with machine learningbased decision support. The system automates price recommendation based on purchase order quantity, product type, region, and predicted seasonal demand, while retaining full administrative control through mandatory price approval. A demand forecasting module using machine learning techniques such as Random Forest regression analyzes historical sales data to support inventory planning and pricing decisions. In addition, an AI-powered chatbot enhances customer interaction by assisting with product queries and navigation. By integrating modern web technologies, AI-assisted pricing, demand forecasting, and CI/CD-based automation, the platform delivers a scalable, intelligent, and business-controlled digital sales solution suitable for real-world deployment.

2. PROBLEM FORMULATION
   1. Core Problem Statement

      Primary Challenge addresses to design and develop an intelligent e-commerce system for a manufacturing business that automates pricing decisions, supports seasonal demand and inventory management, and improves customer interaction, while maintaining full administrative control over critical business operations.

      Specific Problem Components

      Manual and static pricing uses traditional systems that are fixed or manually adjusted prices that do not account for order quantity, regional demand variations, or seasonal trends. This results in inconsistent pricing, reduced competitiveness, and delayed decision- making.

      Lack of demand and inventory intelligence manufacturers often lack tools to predict seasonal demand using historical sales data. As

      a result, inventory planning is reactive rather than proactive, leading to overstocking, understocking, or missed sales opportunities.

      Limited customer support and interaction has conventional e-commerce platforms that provide minimal assistance to customers. Users often struggle to find suitable products or obtain quick answers to queries, negatively impacting user experience and conver- sion rates.

      Operational inefficiency in system updates manual software deployment and testing processes increase development time and in- troduce a higher risk of errors, making it difficult to maintain system reliability and consistency.

      1. 1. Problem Solution
            1. Solution Architecture Overview

               The system follows a modular full-stack architecture in which intelligent backend services support decision-making while the frontend delivers a simple and responsive user experience. Machine learning models operate as decision-support components rather than autonomous controllers, ensuring that final business decisions remain under administrative supervision.

            2. Core Solution Components

               AI-Assisted Pricing Engine: The system includes an AI-assisted pricing engine that generates price recommendations based on product type, purchase order quantity, customer region, predicted seasonal demand, and internally maintained competitor pricing data. The pricing engine functions as a decision-support system rather than a fully automated controller. All generated prices are forwarded to the administrator for review and approval before being published to customers, ensuring transparency and business control.

               Demand Forecasting Module: A machine learningbased demand forecasting module is integrated to predict future demand patterns using historical sales data along with regional and seasonal factors. The module is implemented using a Random Forest regression model and produces a demand index that supports pricing decisions and inventory planning. The predicted demand is used internally by the system and is not exposed to customers.

               Chatbot-Based Customer Assistance: An AI-powered chatbot is incorporated to assist customers with product-related queries, nav- igation, and basic order guidance. The chatbot uses a pretrained natural language processing model to understand user intent and provide relevant responses, improving customer engagement and reducing manual support requirements.

               Administrative Control and Approval Workflow: The system provides a dedicated administrative interface through which adminis- trators can manage products, review pricing recommendations, approve or modify prices, and monitor order activity. This workflow ensures centralized control over all critical business decisions.

               Technical Implementation Strategy

               Pricing Recommendation Pipeline: Pricing recommendations are generated using business-defined rules supported by demand fore- casting outputs and competitor pricing references. The pricing logic is implemented in the backend to maintain separation between user inteaction and business intelligence.

               Machine Learning Pipeline for Demand Forecasting: Historical sales data is preprocessed and used to train a Random Forest regres- sion model. The trained model predicts future demand, and its output is integrated into the pricing engine as a backend-only input.

               Chatbot Processing Pipeline: User queries are processed using a pretrained NLP model for intent understanding and response gen- eration. The chatbot logic is restricted to assistance and does not trigger direct system actions such as pricing changes or order confirmation.

               CI/CD Automation Support: CI/CD practices using GitHub Actions are employed to automate code integration, testing, and deploy- ment, improving development efficiency and system reliability without affecting runtime behavior.

            3. Integration and Security Framework

      The proposed system integrates multiple external APIs and security mechanisms to ensure reliable and secure operation. Payment processing is handled through secure gateway services such as Razorpay or Stripe, enabling safe online transactions. For conversational assistance, chatbot processing utilizes pretrained natural language processing services accessed through OpenAI API or Hugging Face inference APIs to support intent understanding and response generation. Data persistence and management are

      implemented using MongoDB Atlas, a cloud-managed NoSQL database platform that provides scalability and flexible schema support. Security implementation within the system includes JWT-based authentication combined with role-based access control to ensure authorized access to administrative functionalities. All API communications are protected using HTTPS encryption, and sensitive information such as API keys and credentials is securely stored using environment variables to prevent unauthorized exposure. In addition, robust error handling and validation mechanisms are implemented across all integrated services to manage failures in payment processing, chatbot responses, machine learning predictions, and database operations, thereby maintaining system stability and providing reliable user feedback.

3. LITERATURE SURVEY

   Recent research highlights a growing adoption of artificial intelligence and machine learning techniques in e-commerce systems to overcome the limitations of static pricing, manual inventory management, and limited customer interaction. Traditional e-commerce platforms rely heavily on predefined pricing and reactive inventory strategies, which restrict their ability to respond to fluctuating demand and regional market conditions. To address these challenges, researchers have proposed intelligent systems that combine pricing automation, demand forecasting, and decision-support mechanisms.

   Dynamic and AI-assisted pricing has been widely studied as a method to improve revenue optimization and market competitiveness. Chen et al. [1] demonstrate that machine learningbased pricing models can generate adaptive price recommendations by analyzing order quantity, demand trends, and regional factors. Patel and Shah [3] further emphasize the importance of human-in-the-loop pricing systems, where AI provides recommendations while final pricing decisions remain under administrative control. This ap- proach improves transparency and reduces the risks associated with fully automated pricing systems, particularly for small and medium-scale enterprises.

   Demand forecasting plays a critical role in supporting pricing and inventory decisions. Kumar et al. [2] and Verma et al. [4] report that Random Forestbased regression models are effective in predicting seasonal and regional demand patterns using historical sales data. These studies highlight that accurate demand prediction enables businesses to anticipate future requirements and avoid over- stocking or stock shortages. Further advancements by Zhang et al. [11] and Wang et al. [18] show that integrating demand forecast- ing outputs directly into pricing and inventory systems improves overall operational efficiency and decision accuracy.

   Inventory management has also benefited from predictive analytics. Lee et al. [6] and Mehta et al. [14] present systems where forecast-driven inventory planning helps manufacturing and retail organizations optimize stock levels. Their findings suggest that combining demand prediction with administrative oversight results in more reliable and scalable inventory management solutions, particularly in sectors influenced by seasonal demand variations.

   Customer interaction in e-commerce platforms has evolved through the integration of conversational AI technologies. Das et al. [5], Adamopoulou et al. [13], and Kim and Park [19] discuss the use of chatbot-based systems to enhance customer engagement, provide real-time assistance, and simplify product navigation. These studies conclude that chatbots are most effective when deployed as support tools that assist users without directly executing transactional or pricing actions, thereby maintaining system reliability and trust.

   From a system development perspective, modern e-commerce platforms increasingly adopt full-stack web architectures and auto- mated deployment practices. Singh et al. [7] and Fernandez et al. [15] highlight the effectiveness of MERN-based architectures combined with CI/CD automation for building scalable and maintainable web applications. Brown et al. [8] further show that CI/CD pipelines using tools such as GitHub Actions reduce deployment errors and improve development efficiency without impacting runtime behavior.

   Overall, the reviewed literature indicates a clear trend toward intelligent, AI-assisted e-commerce systems that balance automation with human control. While existing studies address pricing optimization, demand forecasting, inventory management, and customer assistance independently, limited work focuses on integrating all these components within a single controlled architecture. The proposed Halchal E-Commerce Platform addresses this research gap by combining AI-assisted pricing, machine learningbased demand forecasting, chatbot-based customer support, and CI/CD-enabled development automation into a unified and business- controlled system.

4. ARCHITECTURE

   Fig. 1. Architecture of System.

   The system architecture of the proposed Halchal E-Commerce Platform is illustrated in Fig. 1. The architecture follows a modular full-stack design that integrates intelligent decision-support components with controlled business workflows. The system is struc- tured to separate user interaction, backend processing, AI-based services, administrative control, and data persistence, thereby en- suring scalability, security, and transparency.

   The frontend layer is developed using React.js and serves as the primary interface for both customers and administrators. Through this interface, users can browse products, manage carts, interact with the chatbot, and initiate order placement, while administrators can access dashboards for reviewing and approving pricing decisions. The frontend communicates with the backend using secure RESTful APIs.

   The backend layer, implemented using Node.js with Express.js, acts as the central controller of the system. It handles authentication, order processing, pricing requests, chatbot query routing, and communication with AI services and the database. All business rules and validations are enforced at this layer to maintain system consistency.

   The AI services layer includes two core components: demand forecasting and AI-assisted pricing. The demand forecasting module uses a Random Forest machine learning model to analyze historical sales data along with seasonal and regional factors. The predicted demand is generated in batch mode and passed as an internal input to the AI-assisted pricing module. The pricing module computes price recommendations based on product type, purchase quantity, customer reion, predicted seasonal demand, and internally main- tained competitor pricing data. These recommendations are not applied automatically.

   An administrative approval layer acts as a control mechanism between the pricing intelligence and data storage. All price recom- mendations generated by the AI-assisted pricing module are reviewed by an administrator. Prices are finalized only after explicit approval, ensuring transparency and preventing uncontrolled automated pricing.

   The system also integrates a chatbot service, implemented using a pretrained NLP model accessed via an external API. Customer queries are routed from the backend to the chatbot service, which returns contextual responses related to products and navigation. The chatbot operates strictly as a support component and does not perform transactional or pricing actions.

   The data layer uses MongoDB Atlas as a cloud-hosted NoSQL database to store user data, product information, orders, approved prices, competitor pricing references, and historical sales records. Payment transactions are processed securely through an external payment gateway such as Razorpay or Stripe, with transaction status updates handled by the backend.

   Additionally, CI/CD pipelines implemented using GitHub Actions support automated build, testing, and deployment processes. These pipelines improve development efficiency and deployment reliability but do not affect the runtime behavior of the system.

   Overall, the proposed architecture effectively combines AI-assisted pricing, machine learningbased demand forecasting, chatbot- based customer assistance, and administrative control within a single unified framework, making it suitable for both academic evaluation and real-world deployment.

5. IMPLEMENTATION DETAILS

   The implementation of the proposed Halchal E-Commerce Platform is organised into multiple interconnected modules designed to support intelligent pricing, demand forecasting, administrative control, customer interaction, and secure transaction management. The AI-assisted pricing module generates price recommendations by analysing product type, purchase quantity, customer region, predicted seasonal demand, and internally maintained competitor pricing data. The generated prices are not directly published to customers; instead, they remain in a pending state until administrative approval is completed. This human-in-the-loop mechanism ensures controlled automation while maintaining transparency and business oversight.

   The demand forecasting module is implemented using a Random Forest regression model that analyses historical sales data, seasonal indicators, and regional demand patterns to predict future demand trends. The forecasting process operates in batch mode and pro- duces a demand index that supports pricing decisions and inventory planning. Since forecast accuracy depends on the availability and quality of historical datasets, the module is designed as a decision-support system rather than a fully automated controller.

   The administrative approval module is incorporated to enable administrators to review, approve, or modify AI-generated price recommendations before publication. This module is integrated within the administrative dashboard and enforces a strict manual approval workflow, ensuring that all finalised prices align with business policies.

   A chatbot assistance module is integrated into the platform to improve customer interaction. The chatbot utilizes a pretrained natural language processing model accessed through an external API to assist users with product-related queries, navigation guidance, and frequently asked questions. The chatbot operates strictly as an informational support component and does not execute transactional or pricing actions.

   The order and payment module manages order placement, payment processing, and transaction status tracking through secure pay- ment gateway integrations such as Razorpay or Stripe. The system ensures secure communication and reliable transaction handling, contributing to a high transaction success rate. Furthermore, an inventory support module leverages forecasted demand insights to assist administrators in seasonal stock planning and inventory awareness. This module functions as a non-automated decision-sup- port tool, allowing administrators to make informed inventory adjustments.

   The backend framework of the system is implemented using Node.js with Express.js to develop modular RESTful APIs that handle authentication, product management, pricing logic, forecasting access, chatbot routing, and order processing. The lightweight server configuration ensures scalability and suitability for academic as well as prototype deployments. On the frontend, the platform utilizes HTML5, CSS3 with Bootstrap 5, JavaScript (ES6), and React.js to create a responsive single-page application interface. The ad- ministrative dashboard provides visual access to pending price recommendations, approved prices, inventory indicators, and order summaries, enabling efficient administrative control. Core interface components include product listing and cart management, checkout and order tracking pages, chatbot interaction panels, and administrative tools for pricing approval and product manage- ment.

   The backend framework of the proposed system is implemented using Node.js with Express.js to develop RESTful APIs that manage authentication, product management, pricing logic, forecasting access, chatbot routing, and order processing. The server configura- tion follows a lightweight architecture suitable for academic and prototype deployment while maintaining scalability and modular- ity. The API design adopts a structured approach that separates business logic from user interaction, ensuring secure communication and efficient data processing across all integrated services.

   The frontend interface of the platform is developed using HTML5, CSS3 with Bootstrap 5, and JavaScript (ES6), with React.js used to implement a responsive single-page application architecture. The user interface supports dynamic chatbot interaction, product browsing, and order tracking while providing administrators with a dashboard to review pending price recommendations, approved prices, inventory indicators, and order summaries. Core interface components include product listing and cart management, checkout and order tracking pages, chatbot UI integration, and an administrative dashboard designed for pricing approval, product manage- ment, and demand overview visualization through RESTful API communication.

   The machine learning implementation within the proposed system focuses on demand forecasting using a Random Forest regression model developed with the scikit-learn library. The model is trained using historical sales data and incorporates multiple input fea- tures including product type, customer region, historical sales quantity, and seasonal indicators to capture demand variations across time and geography. The forecasting process operates periodically in batch mode and produces a predicted demand index that is utilized internally for pricing recommendations and inventory planning support. The generated forecast values are not exposed directly to customers; instead, they function as backend decision-support inputs that enhance system intelligence while maintaining administrative control over business operations. The forecasting model was evaluated using historical validation data to ensure reliable demand prediction accuracy.

   The AI-assisted pricing logic integrates multiple pricing inputs such as base product price, purchase quantity, customer region, predicted seasonal demand, and competitor pricing references maintained by administrators. Based on these factors, the pricing engine generates dynamic price recommendations that support informed decision-making while preserving business transparency. The system operates under a human-in-the-loop framework in which generated prices remain in a pending state until reviewed and approved by an adminisrator. This approach ensures fairness, pricing consistency, and compliance with organizational policies while preventing fully automated pricing decisions from being applied without supervision.

   The chatbot processing module utilizes a pretrained natural language processing model accessed through external APIs such as OpenAI API or Hugging Face Inference API to support intelligent user interaction. The chatbot is designed to assist users with product-related information queries, navigation guidance, and responses to frequently asked questions, thereby improving customer engagement and usability within the platform. The operational scope of the chatbot is limited to informational assistance, and it does not execute transactional operations, pricing modifications, or administrative actions. This controlled design ensures that au- tomated responses enhance user experience while maintaining strict system security and administrative governance.

   The data storage architecture of the system is implemented using MongoDB Atlas, a cloud-hosted NoSQL database that provides scalability, flexible schema design, and seamless integration with the Node.js backend. The database maintains multiple collections including user and role information, product data, order and payment status records, approved pricing details, competitor pricing references, and historical sales datasets used for demand forecasting. MongoDB is selected to support dynamic data structures and efficient query performance, enabling reliable storage management and facilitating smooth communication between machine learn- ing components, backend services, and administrative dashboards.

   The system incorporates multiple security and authentication mechanisms to ensure safe access and data protection across all plat- form components. User and administrator authentication is implemented using JWT-based authorization combined with role-based access control to restrict sensitive administrative operations. All API communications are secured through HTTPS encryption to prevent unauthorized data interception during transmission. Additionally, sensitive credentials such as API keys and configuration secrets are securely stored using environment variables, ensuring compliance with secure development practices and protecting critical system resources from exposure.

   The platform integrates CI/CD automation using GitHub Actions to streamline development, testing, and deployment processes. The continuous integration pipeline manages code integration and automated checks to maintain code quality and detect potential issues early in the development lifecycle. Following successful validation, the deployment stage enables seamless application up- dates while minimizing downtime and maintaining system stability. This automated workflow enhances development efficiency, ensures consistent software delivery, and supports reliable deployment practices aligned with modern DevOps methodologies.

   Fig. 2. Admin Dashboard

6. CONCLUSION

This paper presented an AI-enabled e-commerce platform for Halchal Industries that integrates AI-assisted pricing, machine learn- ing-based demand forecasting, chatbot support, and CI/CD automation within a unified full-stack architecture. The proposed system generates intelligent pricing recommendations based on purchase quantity, regional factors, and seasonal demand while maintaining administrative control through a mandatory approval workflow. The Random Forest-based forecasting model supports informed pricing and inventory planning, and the chatbot enhances customer interaction without executing transactional operations. The overall architecture demonstrates a scalable and practical approach to integrating artificial intelligence with controlled business workflows, making it suitable for real-world deployment as well as academic evaluation.

Future work may include real-time adaptive pricing using reinforcement learning and live demand-stream integration.

REFERENCES

1. Machine LearningBased Dynamic Pricing Models for E-Commerce Platforms Chen et al., 2023
2. Demand Forecasting Using Random Forest for Retail and Manufacturing Systems Kumar et al., 2023
3. AI-Assisted Pricing Systems with Human-in-the-Loop Control Patel and Shah, 2023
4. Seasonal Demand Prediction for Inventory Optimization Using Machine Learning Verma et al., 2023
5. Chatbot-Driven Customer Engagement in E-Commerce Applications Das et al., 2023
6. Intelligent Inventory Management Using Predictive Analytics Lee et al., 2023
7. Secure and Scalable E-Commerce Platforms Using MERN Stack Singh et al., 2023
8. CI/CD Automation for Web Applications Using GitHub Actions Brown et al., 2023
9. Machine Learning Models for Sales Forecasting in Online Markets Taylor and Wilson, 2023
10. Ethics and Transparency in Algorithmic Pricing Systems Ezrachi et al., 2023
11. AI-Driven Demand Forecasting for Supply Chain Optimization Zhang et al., 2024
12. Decision-Support Pricing Systems Using Machine Learning Roberts and Lee, 2024
13. Conversational AI Chatbots for Online Retail Platforms Adamopoulou et al., 2024
14. Predictive Analytics for Inventory Planning in Manufacturing Industries Mehta et al., 2024
15. Cloud-Based E-Commerce Systems with Automated Deployment Pipelines Fernandez et al., 2024
16. Human-Centered AI for Business Decision-Making Systems Rahwan et al., 2024
17. Machine LearningEnabled Pricing Optimization for SMEs Oliveira et al., 2024
18. Demand Forecasting and Pricing Integration in Smart Commerce Systems Wang et al., 2025
19. AI-Assisted E-Commerce Platforms with Chatbot Integration Kim and Park, 2025
20. Automated Pricing and Inventory Control Using Predictive Models Gupta et al., 2025