Innovation Portfolio

AI, Assistive Systems & Digital Platforms

Real-Time Video De-Hazing and Visibility Enhancement System

Dense fog, smoke, and haze reduce visibility and make object identification difficult in surveillance, rescue, and navigation environments. This system enhances real-time video to restore clarity and enables detection of objects even under low-visibility conditions.

Healthcare & Biomedical Systems

AI Powered Primary Healthcare System

The healthcare industry has gone through major technological advancements over the past ten years, with Artificial Intelligence (AI) becoming one of the key technologies supporting medical applications. AI-based systems are increasingly used to analyse information, assist healthcare professionals, and improve access to medical knowledge. These developments have created new opportunities to build intelligent systems that can provide quick and reliable information while supporting medical education and clinical decision-making. Traditional methods of searching medical textbooks are often time-consuming, particularly when users need immediate access to information related to symptoms or diseases. Students and healthcare professionals usually spend considerable time locating relevant sections within large medical textbooks. An intelligent search system that understands the meaning of a user's query rather than relying only on keywords can significantly improve this process. The AI-Powered Primary Healthcare Consultation System was developed to address this challenge. Instead of functioning as a medical diagnosis tool, the system acts as an educational healthcare assistant by retrieving information directly from Davidson's Principles and Practice of Medicine. It combines semantic search with a Large Language Model (LLM) to generate responses that are supported by the textbook while displaying the corresponding page references. This approach allows users to verify the information instead of relying on generated text alone. The system suggested uses the Retrieval-Augmented Generation method. The medical textbook is turned into digital text through the use of Optical Character Recognition (OCR). The extracted content is arranged into a structured dataset and turned into vector embeddings with the help of Sentence Transformers. Facebook's AI tool, called FAISS, is used to find the most relevant parts of textbooks. Then, Google's Gemini model uses those found parts to create a short and clear answer. The full application is set up using Streamlit, which gives users an easy-to-use interface to interact with. The project demonstrates how modern AI technologies can improve access to medical knowledge while maintaining transparency by linking every response to its original textbook source. Rather than replacing healthcare professionals, the system serves as an educational tool that assists users in locating reliable medical information efficiently.

ARNAV THAPLIYAL

AI, Assistive Systems & Digital Platforms

Real-Time Posture Correction System for Sport Personnel

Incorrect body posture during daily activities, sports, and training can lead to injury and long-term musculoskeletal problems. This system monitors body posture in real time and provides immediate corrective feedback, helping users maintain proper alignment and improve movement safety and performance.

IoT, Energy & Smart Infrastructure

Wireless Smart Streetlight Monitoring and Control System

Traditional streetlight systems rely on manual inspection, causing delayed fault detection, energy loss, and safety risks. This system enables real-time monitoring and automatic fault detection of streetlights, ensuring efficient operation and improved public safety. It provides remote control, precise fault localization, and reliable communication without internet dependency, making it suitable for smart and sustainable urban infrastructure.

IoT, Energy & Smart Infrastructure

Augmented Reality Based Road Tracking and Navigation System

Operating heavy machinery in snow, fog, desert, or disaster environments is difficult due to poor visibility and lack of clear road guidance. This system provides real-time augmented visual navigation, helping operators track road alignment, turns, slopes, and obstacles safely. It improves situational awareness, reduces accident risk, and enhances operational efficiency in challenging terrains and low-visibility conditions.

Healthcare & Biomedical Systems

Smart Medicine Reminder and Compliance Monitoring System

Missing or incorrect medication timing can lead to serious health risks, especially for elderly and long-term patients. This system ensures timely medicine intake through intelligent reminders, compartment-level alerts, and real-time monitoring.It improves medication adherence, supports caregivers through remote alerts, and maintains safe storage conditions for reliable healthcare assistance.

AI, Assistive Systems & Digital Platforms

Interactive Legal Classification and Statutory Mapping System

Identifying correct legal sections from incident descriptions is difficult for non-legal users and often leads to misclassification or delays. This system analyzes user-described incidents and automatically maps them to relevant statutory provisions, providing instant legal guidance. It improves legal accessibility, supports faster preliminary assessment, and helps ensure accurate classification of incidents in real-world scenarios

Healthcare & Biomedical Systems

IoT-Based Smart Medicine Dispenser

The IoT-Based Smart Polypharmacy Medicine Dispenser is an intelligent healthcare device designed to help patients take multiple medicines on time and reduce medication errors. It uses an ESP32 microcontroller, DS3231 Real-Time Clock (RTC), DC motors, and a 3D-printed enclosure to automate medicine dispensing. Each medicine is stored in a separate drawer and can be scheduled through a web application. The ESP32 continuously checks the programmed schedule and dispenses the correct medicine at the right time. Firebase Realtime Database enables real-time communication between the hardware and the web application for remote monitoring and control. The TM1637 display shows the current time and the drawer number during medicine reminders. The system also allows caregivers to monitor and manage medication schedules remotely. Overall, the project provides an affordable, reliable, and user-friendly solution that improves medication adherence and supports smart healthcare.

PARAMJEET SINGHKEERTI SINGH

AI, Assistive Systems & Digital Platforms

Realtime Football Curve Shot Posture Assessment Using Markerless Pose Estimation

The Real-Time Football Curve Shot Posture Assessment System is an AI-enabled sports analytics innovation developed to evaluate a player’s body posture during a football curve shot using multi-view video analysis and markerless pose estimation. The system is designed to support coaches, players, trainers, and sports researchers by providing an objective, explainable, and technology-driven method for posture assessment without requiring wearable sensors or expensive motion-capture equipment.The system captures the player’s kicking action from multiple camera angles, such as front, side, and rear views. These video streams are synchronized using an audio-based clap signal, allowing the system to analyze the same moment of action from different perspectives. The key event in the shot, the exact foot-ball contact frame, is identified by tracking the ball movement and the position of the player’s ankle.Once the contact frame is detected, the system applies markerless pose estimation to extract important body keypoints, including the shoulder, hip, knee, ankle, and foot. These keypoints are then used to calculate biomechanical posture parameters such as knee angle, trunk angle, and ankle angle. The calculated posture values are compared with predefined assessment ranges to generate rule-based feedback and an overall posture score.This innovation aims to make sports performance analysis more accessible, practical, and interpretable. It can help players understand their body alignment, balance, joint positioning, and technique during a curve shot. For coaches, the system provides a visual and data-supported method to identify posture errors and guide performance improvement.The technology represents a step toward low-cost, AI-based sports biomechanics solutions that can be used in training environments, academic research, sports coaching, and performance evaluation. By combining computer vision, artificial intelligence, and biomechanical analysis, the system demonstrates how modern digital tools can contribute to smarter and more scientific sports training.Key FeaturesMulti-view football shot video analysisAudio-based video synchronizationMarkerless pose estimation without wearable sensorsAutomatic foot-ball contact frame detectionKnee, trunk, and ankle angle calculationRule-based posture evaluationVisual feedback dashboard and posture scoringPractical application in sports training and biomechanics researchPotential ApplicationsFootball coaching and player trainingSports biomechanics researchAI-based performance assessmentInjury-risk awareness and posture correctionAcademic research and product-based innovationDevelopment of low-cost sports analytics toolsDeveloped under the innovation and research ecosystem of Amity Innovation and Design Centre (AIDC), Amity University Uttar Pradesh, this system reflects AIDC’s focus on transforming interdisciplinary research ideas into practical, product-oriented solutions.

Rahul YadavDr. Vinayak MajhiDr. Gurpreet KaurDr. Soni GuptaDr. Sujata Pandey

AI, Assistive Systems & Digital Platforms

Adaptive Threat Detection and Secure Data Transmission in Edge Based Healthcare IoT Systems

This project presents two complementary pieces of work carried out in the domain of embedded systems and Internet of Things (IoT) security, representing a progression from a foundational hardware prototype to an advanced, research-grade security framework. The first system, MoodBot: Robot Desk Buddy, is an Arduino UNO-based embedded desk companion that continuously senses ambient light intensity through a Light Dependent Resistor (LDR) and reports real-time lighting conditions — "BRIGHT", "DIM", or "DARK" — on a 16×2 I2C LCD display. It addresses the everyday problem of poor desk lighting that causes eye strain and reduced productivity for students, office workers, and hobbyists learning embedded systems, and establishes core competencies in sensor interfacing, I2C communication, and real-time feedback loops. Building on this embedded-systems foundation, the second and principal contribution of this report, MediShield, is a severity-aware adaptive security framework for energy-efficient healthcare IoT data protection. Unlike conventional wearable health monitors that apply a single strong cipher — typically AES-256 — to every physiological record regardless of clinical significance, MediShield derives a composite risk score (0–12) from four vital parameters (heart rate, blood-oxygen saturation, body temperature, and respiratory rate), classifies each incoming record into one of four severity tiers (Normal, Mild, Moderate, Critical), and binds each tier to a proportionate encryption algorithm (None, SPECK 64/128, AES-128, AES-256) and transmission policy. Evaluated on a labeled dataset of 25,493 real physiological records and demonstrated end-to-end through a browser-based dashboard, an ESP32 battery drain simulator, and a Python/Raspberry Pi sensor relay streaming data through a Google Sheets cloud pipeline, MediShield reduces cumulative encryption overhead by 88.96%, cloud transmissions by 46.41%, and total energy consumption by 65.52% — equivalent to nearly 2.9× longer battery operating life — while guaranteeing that every Critical-tier record continues to receive AES-256 protection. The severity classification engine achieves 100% rule-validation accuracy across the full dataset. Taken together, MoodBot and MediShield illustrate a coherent development trajectory: from a simple, sensor-driven embedded prototype that builds fundamental hardware and firmware skills, to a rigorously evaluated, clinically motivated adaptive security framework that addresses a real and pressing problem in resource-constrained healthcare IoT deployments.

Ishika GuptaSiddh SharmaRaj Ranjan

AI, Assistive Systems & Digital Platforms

Neural World Modern Agent

The Neural World Model Agent (NWMA) is a multi-agent artificial intelligence system designed tosupport decision-making across four heterogeneous business domains: Finance, Marketing,Economics, and Supply Chain. The system integrates an Encoder-World Model-ReinforcementLearning (RL) Controller architecture with a Random Forest classification layer to transform rawdomain data into actionable, interpretable decisions. Each domain is represented by a dedicatedinput agent that feeds structured features into a shared encoder, which compresses high dimensional observations into a compact latent representation. A learned world model thenpredicts how this latent state evolves under candidate actions, allowing the RL controller to planand select policies without exhaustively interacting with the real environment. The resulting policysignal, combined with engineered domain features, is passed to a Random Forest classifier thatproduces the final categorical decision output, such as risk classification, customer segment,demand category, or pricing action. Across all four domains, the proposed architecture achieved an average classification accuracy of 90.65%, an improvement of approximately 8.9 percentage points over a standalone Random Forest baseline that does not incorporate the encoder-world-model-RL pipeline. The RL controller's reward curve exhibited stable convergence after approximately 120 training episodes, and feature importance analysis confirmed that the latent representations learned by the encoder contributed morepredictive value than raw domain features alone. These results indicate that a shared world-model based architecture can generalize effectively across structurally different business domains while remaining computationally lightweight enough for practical deployment. The report further discusses the theoretical grounding of the approach in reinforcement learning and model-based control literature, positions the work relative to existing domain-specific machine learning systems,and outlines directions for future work, including transformer-based encoders and real-time deployment pipelines.

AI, Assistive Systems & Digital Platforms

Smart Women Safety Application

SheSafe is a real-time women safety response application designed to provide immediate emergency assistance during critical situations. The application enables users to send SOS alerts along with their live location to a Police Control Dashboard using Firebase Realtime Database. The system is developed using HTML, CSS, JavaScript, Browser Geolocation API, and Google Maps. It provides a fast, reliable, and cloud-based emergency response solution.

AI, Assistive Systems & Digital Platforms

AI Adaptive Placement Preparation & Learning Intelligence Ecosystem

<span lang="EN-US" style="font-size:12.0pt;line-height:107%; font-family:"Times New Roman",serif;mso-ansi-language:EN-US">AI Adaptive Placement Preparation & Learning Intelligence Ecosystem was created to help students in having a customized and effective learning process in preparation for placements. This system makes use of the concepts of adaptive learning, performance management, AI recommendations and assessment.

IoT, Energy & Smart Infrastructure

Physical Smart Grid Energy Trading

<span style="font-size:11.0pt;font-family:"Times New Roman",serif; mso-fareast-font-family:"Times New Roman";color:black;mso-ansi-language:EN-IN; mso-fareast-language:EN-GB;mso-bidi-language:AR-SA">The objective of this project is to design and develop a Physical Smart Grid Energy Trading Diorama that demonstrates the generation, storage, monitoring, and sharing of electrical energy using renewable energy sources.

PRINCE PRATAP SINGH

AI, Assistive Systems & Digital Platforms

Advanced Education Content Platform

Advanced Educational Platform is a web-based application designed to provide an interactive, user-friendly, and efficient digital learning environment for students and educators. The platform aims to overcome the limitations of traditional learning systems by offering centralized access to educational resources, structured courses, and progress tracking through a modern and responsive interface. The application enables students to register, log in securely, browse available courses, enroll in learning programs, access educational content, and monitor their learning progress. Administrators can efficiently manage users, courses, educational materials, and other platform-related activities through a dedicated administration panel. The system emphasizes ease of use, accessibility, and an improved user experience while ensuring secure authentication and effective data management. The project has been developed using modern web technologies, including HTML, CSS, JavaScript, Java, JSP/Servlets, Apache Tomcat, and MySQL. These technologies provide a reliable, scalable, and maintainable architecture capable of supporting future enhancements and additional functionalities. The primary objective of this project is to create a comprehensive educational platform that simplifies online learning, enhances student engagement, and enables efficient management of educational resources. The proposed system offers a practical solution for educational institutions and online learning environments by integrating essential learning features into a single, centralized platform.

Drone, Robotics & Advanced Engineering

Implementation of a LiDAR-Based Autonomous Campus Delivery Robot using ROS2 and SLAM

This project aims to create and deploy an Autonomous Mobile Robot (AMR) that can move around changing, non-structured campus settings for effective last-mile delivery. The robot system features a decentralized, dual-processing setup where complex spatial programming and hardware control are separately handled.

AI, Assistive Systems & Digital Platforms

ORCHESTRA AI: Music Enhancer and AI Orchestra Generation System

The ORCHESTRA AI: Music Enhancer and AI Orchestra Generation System is an Artificial Intelligence-based software application developed to simplify the process of digital music enhancement by automatically generating synchronized orchestral accompaniments while preserving the original characteristics of the uploaded music. The primary objective of the system is to reduce manual effort involved in conventional music production by combining intelligent audio analysis, AI-assisted orchestration, and an easy-to-use graphical interface into a single integrated platform.

Sanyam Sharma

Healthcare & Biomedical Systems

Health Polymedic

ANMOL GABARISHABH KAPOORADARSH KAMAT