SIVASRIRAMAN P

I am Sivasriraman P who is deeply passionate about data science, machine learning, and data structures, continually seeking opportunities to expand my knowledge in these fields. I thrive on solving complex challenges, particularly those related to graph-based problems, and frequently engage with platforms like Codeforces to enhance my problem-solving abilities.

My dedication to mastering the intricacies of algorithms and machine learning models and applying them to real-world scenarios is reflected in my consistent efforts to grow and improve in these dynamic areas of computer science. I am also focused on enhancing my expertise in image processing, deep learning (DL), and full-stack development, ensuring a well-rounded skill set.

Medico 2025 — VQA for Gastrointestinal Imaging
• Fine-tuned PaliGemma 2 for Visual Question Answering on GI endoscopy images.
• Improved interpretability using Grad-CAM visual explanations and confidence estimation.
• Enhanced clinical reliability through multimodal reasoning and structured outputs.
Real-Time Deepfake Detection System
• Built an AI-based system to detect deepfakes in uploaded videos and live streams with confidence
scoring.
• Used deep learning models with OpenCV and TensorFlow/PyTorch for real-time inference.
• Integrated a blockchain framework for secure, immutable storage of high-confidence predictions.
Fake Product Detection in Sustainable Goods Market
• Developed a hybrid AI + Blockchain solution to detect misleading sustainability claims on e
commerce platforms.
• Applied NLP (BERT) and ML models (Random Forest, Gradient Boosting) to analyse product
data.
• Verified eco-certifications via blockchain; generated dynamic credibility scores to flag greenwashing.
Multilevel Political Meme Classification
• Modelled the task as a multimodal classification problem combining image and text features.
• Used ConvNeXt-Tiny for visual encoding and XLM-RoBERTa with OCR-extracted text embed
dings.
• Applied fusion networks with dual classification heads and class-weighted training for macro-F1
optimisation.
Customizable Game Simulator (WebSockets-Based)
• Built a multiplayer 2-player game platform with private rooms and real-time communication.
• Used React (frontend), Django Channels (WebSockets), and Flask for backend services.
• Implemented JWT authentication and customisable rules/board configurations.
EzCabz — Cab Booking Application
• Developed a real-time cab-hailing mobile application using Python and Kivy.
• Implemented booking, driver tracking, fare estimation, and feedback modules.
• Integrated Django REST API, Flask-SocketIO, SQLite, and SMTP for backend and notifications.
Course Enrollment System — Spring Boot
• Developed a web-based course enrollment system using Spring Boot in Java.
• Designed the frontend in HTML/CSS with SQLite database integration.
• Implemented student registration, course selection, and enrollment management modules.
Magnetic Intensity Controller — AI-Based Control System
• Developed an ML model to predict electric current required for lifting metallic objects via electro
magnet.
• Used object weight and shape as input features; validated against magnetic force equations.
• Designed for integration into a real-time feedback control system for safe industrial lifting.