Cloud Native Specialist | AI & ML enthusiast | GenAI Developer | Innovator

Hima V S  • Peerlist

Implemented and optimized Blaze Pose, a lightweight convolutional neural network architecture, for human pose estimation on mobile devices. Investigated the application of L2 Normalization to rendered coordinates in pose estimation to ensure consistency and accuracy in vector representation. Developed a method for Yoga Pose Assessment using Blaze Pose and DTW to provide real-time feedback on pose correctness. Achieved a frame rate of over 30 frames per second for pose estimation, making it suitable for real-time applications such as fitness tracking. Proposed a novel approach for Yoga Pose Recognition targeting both beginners and experienced specialists. During this research gained strong understanding convolutional neural networks for pose estimation proficiency in TensorFlow, and OpenCV. Above work was presented to the examiners during the course of Bachelor’s of Engineering.

Yoga Pose Assessment System for Self-Learning

Developed an IoT-enabled water metering system featuring LED indicators that illuminate upon the flow of 1 litre of water, providing real-time consumption data. Engineered an ON/OFF mechanism utilizing a photosensitive sensor to accurately detect meter dial rotation, ensuring high-fidelity data collection. Utilized an Arduino Uno microcontroller for streamlined data processing, enabling seamless integration and manipulation of consumption metrics. Established a robust data recording protocol for daily water consumption tracking, facilitating monthly analysis and generation of detailed consumption reports. Implemented a proactive leakage detection mechanism leveraging GSM technology to promptly alert users to anomalies, minimizing water loss and damage. Utilized GSM modules for monthly consumption data dissemination and immediate leakage notifications, ensuring user awareness and intervention capabilities. This work was presented at 3rd International Conference On Innovative Computing And Communication (ICICC-2020), Delhi

Smart Meter

A database management system designed to ease the task of storage and retrieval of driving school-related information. The database maintains information about trainee and trainer details, trainer allotment, vehicle, shift, and fee-related details. Built using Python's Tkinter and integrated with SQLite database

Driving School Management System (DSMS)

A simple platform for farmers to sell their products online. This project is built using HTML, CSS, JavaScript, and PHP. MySQL is used as database server.

E-Commerce Website - Farm Market

Implemented a semantic search engine integrating tools such as Confluence, GitHub, and other workplace tools to streamline project management processes. Developed algorithms to process natural language queries and retrieve relevant information from project data stored in various information hubs. Utilized embedding models including OpenAI's text-embedding-ada-002-v2 (ada-v2) and Google's gcp-textembedding-gecko-001 (gcp-gecko) to convert extracted data into embeddings. Designed and implemented a chat interface for the semantic search engine, enabling users to interactively access project data. Collaborated with AI agents to optimize search algorithms and improve decision-making in query processing. During this research enriched my understanding with GenerativeAI, NLP, embedding models, vector databases. Presented the above work in DemoJam challenge in SAP Dcom-2024

Semantic Search Engine

A Python application that uses the OpenCV library to track an object of interest. It is an implementation of paint application by real-time webcam data.