My Creative Works
A collection of projects I'sve built with passion and precision.
Food Recipe Generator (FYP)
The Food Recipe Generator is a web-based application designed to revolutionize the cooking process by enabling users to generate personalized recipes through the upload of dish images, leveraging advanced artificial intelligence (AI) technologies. Developed as part of my final year undergraduate project in Computer Science at [Your University Name], this project addresses the limitations of traditional text-based recipe platforms by introducing a visual-first approach, making it accessible and intuitive for users of varying cooking expertise. The primary objective was to simplify meal preparation by integrating image-based dish recognition, recipe generation, and practical support features, culminating in a system that enhances user experience in a digital cooking environment. The system’s core functionality relies on a custom-trained image recognition model, utilizing OpenAI's Contrastive Language-Image Pretraining (CLIP) to generate 512-dimensional embeddings and FAISS (Facebook AI Similarity Search) for efficient similarity search against a precomputed dataset of 500 food image embeddings stored in \texttt{food_embeddings.npy}. This model, trained on a diverse set of common dishes (e.g., pasta, pizza, curry) and validated with 50 test images, achieved a recognition accuracy of [insert accuracy]% with an average processing time of [insert time] seconds per upload. Once a dish is identified, the OpenAI API generates detailed recipes, tailored to the visual input, with an average response time of [insert OpenAI response time] seconds. To enhance functionality, the system integrates additional APIs: the Nutritionix API provides nutritional data for downloadable PDF reports (95% query success in [insert Nutritionix response time] seconds), the YouTube API offers relevant cooking video recommendations (90% relevance in [insert YouTube response time] seconds), and the Pexels API supplies high-quality supplementary images ([insert Pexels response time] seconds). A distinctive feature allows users to generate on-demand shopping lists in PDF format, extracted from recipe ingredients via OpenAI and formatted using the FPDF library. The application is architected with a FastAPI backend for robust API management, handling endpoints such as \texttt{/upload-image/}, \texttt{/generate-shopping-list/}, and \texttt{/recipe-generator-history/}, with JWT-based authentication ensuring secure user sessions. The frontend, built with Next.js, delivers a responsive interface styled with TailwindCSS and animated with Framer Motion, providing seamless navigation across devices. PostgreSQL serves as the database, storing user data, interaction logs, recipe searches, PDF records, and ingredient lists in a structured schema with foreign key relationships, optimized with indices for performance. The development process followed an agile methodology, with iterative sprints informed by user feedback from 10 participants, who rated the system’s usability at 4.2/5 and appreciated features like shopping lists (70% positive feedback), though suggesting dietary filters. Testing across 100 end-to-end cases resulted in a 92% success rate, with errors primarily due to dataset limitations (8%) and API timeouts (5%), particularly with YouTube. The project’s contributions lie in its innovative integration of computer vision, natural language processing, and web development, offering a scalable framework for personalized cooking assistance. Compared to platforms like Allrecipes and Yummly, which rely on manual searches, or IBM's Chef Watson, which lacks visual input, this system provides a unique solution with its image-based approach and integrated features. However, limitations include the dataset’s coverage of common dishes only, CPU-based processing constraints, and the absence of multilingual or dietary customization, suggesting areas for future enhancement.
Online Attandance System
The Online Attendance System (OAS) is a web-based application designed to automate and streamline attendance management for organizations, replacing inefficient manual processes with a secure, scalable, and user-friendly digital solution. Developed using the MERN stack (MongoDB, Express.js, React, Node.js), the OAS addresses the needs of modern workplaces by enabling administrators to efficiently record employee attendance and generate reports, while employees can access their attendance records in real-time. The system aims to enhance productivity, reduce errors, and provide actionable insights through attendance data, making it an ideal tool for businesses and institutions seeking cost-effective administrative solutions.
Salman's Custom GPT
I have integrated a custom GPT model into my portfolio project to provide interactive and personalized information about my skills, experience, and projects. This AI-powered assistant can answer queries related to my background in full-stack development, my certifications, and the various technologies I’ve worked with. The GPT model was designed and trained to ensure accurate and contextually relevant responses, making the user experience engaging and informative.
Little Lemon Restaurant
The Little Lemon Restaurant website is a vibrant and inviting online presence for a culinary establishment that specializes in pasta and gourmet dishes. Developed using React and TailwindCSS, this project features a highly responsive design that caters to both desktop and mobile users. The site is deployed on Vercel, ensuring fast loading times and reliable uptime for potential customers checking out menus, specials, or booking a table.
Budget APP
Company's Budget Allocation is a dynamic web application designed to help businesses manage and track their budget distribution across different departments. Developed using React for its efficient rendering capabilities and TailwindCSS for its flexible styling, the app provides a user-friendly interface that is both responsive and intuitive. It is deployed on Vercel for enhanced performance and scalability, ensuring smooth and reliable access.
Seka Solution for Software House
SekaSolution is a cutting-edge web platform designed to showcase various IT services and solutions tailored for global business needs. Built using React and styled with TailwindCSS, this project demonstrates a clean, modern, and responsive front-end interface optimized for both mobile and desktop environments. The website features a streamlined user experience with a professional design that includes interactive elements and a vibrant layout.