Mohammed Farhan I

Projects

About

I’m Farhan, a final-year Computer Science and Engineering (Data Science) student at Mar Athanasius College of Engineering. I’m all about diving into data, building cool software, and exploring web tech.Recent projects I’ve worked on include Steth.Ai for detecting abnormal heartbeats, Slidify for auto-generating slides using LLMs and web scraping, and Vidhya.AI, a smart personal tutor

As a Computer Science Intern at Thakshashila R&D, I collaborated with cross-functional teams to create impactful, real-world solutions. Leveraging my skills in programming, data analysis, and web development, I'm motivated to tackle new challenges and contribute meaningfully to the tech landscape.

Resume

The Steth.AI project aims to develop an AI-integrated stethoscope that enhances the detection and diagnosis of heart, lung, and abdominal conditions by providing real-time feedback. The device captures auscultation sounds through a traditional stethoscope, converts them into digital signals, and processes these using a classical machine learning approach utilizing spectrograms. Key features include abnormality detection, heart disease classification, and real-time analysis on a Wio Terminal using TinyML.
Abnormality Detection: The system analyzes heart, lung, and abdominal sounds in real time, detecting conditions such as heart murmurs, arrhythmias, and lung abnormalities.
Heart Disease Classification: It predicts whether a condition is normal, borderline, or abnormal and classifies common heart diseases, including Aortic Stenosis (AS), Mitral Regurgitation (MR), Mitral Stenosis (MS), and Mitral Valve Prolapse (MVP), alongside normal heart conditions.
Wio Terminal & TinyML Processing: The auscultation data is processed on the Wio Terminal using TinyML, ensuring on-device predictions without the need for cloud processing, making it efficient and accessible in real-time.
The goal of the project is to improve early diagnosis by providing healthcare professionals with an AI-powered tool that enhances traditional stethoscope functionality. Steth.AI is designed to be a user-friendly, portable, and accurate solution for real-time health monitoring, aiding in early detection and better patient outcomes.

The Vidya AI project aims to transform K-12 education by offering personalized, AI-driven tutoring that adapts to each student’s learning style and academic level. Through tailored educational content and adaptive testing, Vidya AI provides a scalable and cost-effective solution to address the current challenges of high tuition fees, limited adaptive learning, and the need for individualized educational support.
Personalized Learning: Vidya AI collects student information such as age, class, and interests to create a tailored academic profile. Based on initial assessments, it customizes teaching content to match each student’s comprehension level, ensuring clarity and a deeper understanding of concepts.
Adaptive Testing: Following each lesson, adaptive tests identify areas where the student needs additional support. Vidya AI uses these insights to adjust teaching methods and reinforce difficult concepts, making learning more effective and engaging.
AI-Powered Video Lessons: Leveraging Language Learning Models (LLMs) and personalized scripts, Vidya AI generates custom teaching videos for each student. The AI adapts the video content to the student’s level of understanding, creating a relatable and highly engaging learning experience.
Vidya AI’s goal is to make high-quality, customized education accessible to all by delivering a flexible, affordable, and student-centered learning experience. This innovative tool enables students to build essential academic and practical skills in a way that is both engaging and effective.

The Slidify project addresses the increasing need for efficient presentation creation by automating the process through the integration of a Language Model (LM) and web scraping techniques. Designed to save time and effort, Slidify takes a text prompt as input and generates coherent, informative content for presentation slides, catering to professionals and educators who seek engaging, visually appealing presentations.
Automated Content Generation: Slidify uses a Language Model to analyze the input text and generate relevant slide content. This capability allows the system to create coherent and contextually accurate slides, streamlining the presentation creation process significantly.
Image Enhancement with Web Scraping: To improve the visual appeal, Slidify employs web scraping to gather relevant images, enhancing the presentation's effectiveness in conveying information through both text and visuals.
Flexible Output Options: Users can specify the number of slides and choose from multiple output formats, including PowerPoint (PPT), LaTeX, and Markdown. The default theme ensures a consistent visual style, while the option to download in various formats adds user convenience.
Slidify’s goal is to streamline presentation creation by offering an intuitive, automated solution that saves time and produces professional-quality presentations. This innovative tool is ideal for anyone needing to communicate effectively and efficiently through well-crafted presentations.

The DocClustr project focuses on the application of document clustering techniques to improve information retrieval and content organization by identifying similar documents within a dataset. Through efficient data categorization, DocClustr enables users to organize and retrieve relevant information more effectively.
Text Processing and Vectorization: The project begins with the extraction and preprocessing of text from documents, transforming them into structured data. Techniques such as TF-IDF vectorization and cosine similarity are used to represent documents based on their content, ensuring an accurate basis for clustering.
K-Means Clustering: The primary clustering method employed is K-Means, chosen for its adaptability and effectiveness in handling various dataset characteristics. Users can adjust the number of clusters to align with the specific structure and needs of the data.
Data Visualization: To enhance result interpretation, DocClustr provides visualizations, including 2D cluster plots and a pie chart showing document distribution. These visual aids help users better understand the clustering outcomes and data structure.
DocClustr offers a comprehensive exploration of document clustering techniques, providing a detailed implementation and insights into the strengths, limitations, and applications of K-Means clustering. This project contributes valuable understanding of clustering approaches and their practical applications in information retrieval.

The Mace Connect project is a Campus Event Approval Platform designed to modernize and streamline college event management. By integrating technology, Mace Connect aims to simplify event approval workflows, minimize delays, and establish a unified platform accessible to students, organizers, faculty advisers, and administrators, fostering a more dynamic college event ecosystem.
Streamlined Approval Process: Mace Connect addresses the common delays and challenges in college event planning by providing a structured, efficient approval workflow. This reduces prolonged waiting periods and minimizes the effort required from organizers, faculty advisers, and administrators.
Centralized Information Hub: The platform serves as a central repository for all event-related information, enhancing accessibility for all stakeholders and allowing for real-time updates and transparent communication within the college community.
Enhanced Community Engagement: By creating an easily accessible platform for event information, Mace Connect promotes a more engaging and active campus environment, allowing students and organizers to focus on event planning and participation rather than administrative hurdles.
Mace Connect envisions an efficient, technology-driven approach to college event management. By reducing manual efforts and improving access to event details, this platform contributes to a more vibrant, accessible, and streamlined experience for college events, benefiting all members of the campus community.

The Relief Hub project is an innovative app designed to enhance transparency in crowdfunding spending management. By leveraging blockchain technology, Relief Hub enables donors to track and vote on fund usage, ensuring their contributions are allocated responsibly and transparently. Built during the Build-on-Chain Hackathon at NIT Calicut during the IEDC Summit, this app addresses key challenges in crowdfunding accountability.
Unique Donor Identification: Each donor receives a unique ID upon contributing to a crowdfund, representing their stake and enabling them to participate in fund management decisions.
Fund Request System: When funds are needed for a specific purpose, recipients must submit a detailed request specifying the amount required and intended purpose. This information is then shared with all donors holding a valid unique ID.
Donor Voting Mechanism: All eligible donors are notified of fund requests and can vote on whether the funds should be released. If over 50% of donors participate and a majority supports the request, the proposal is approved.
Blockchain Transparency: Utilizing Tezos blockchain, Relief Hub records all transactions, fund requests, and voting results, ensuring transparency and immutability of records.
Relief Hub redefines crowdfunding by introducing a transparent, community-driven approach to fund allocation. By allowing donors to have a direct say in fund distribution, this platform fosters trust and accountability in crowdfunding, contributing to a more transparent and responsible ecosystem.