Spearheaded the development of an advanced virtual office assistant using Large Language Models (LLMs), resulting in a 30% increase in operational efficiency through innovative RAG chatbot technology and semantic search algorithms.Engineered cutting-edge machine learning models for automated cricket shot analysis, enhancing player performance feedback accuracy by 40% using state-of-the-art computer vision and deep learning techniques.Led a high-impact collaborative project with… Show more Spearheaded the development of an advanced virtual office assistant using Large Language Models (LLMs), resulting in a 30% increase in operational efficiency through innovative RAG chatbot technology and semantic search algorithms.Engineered cutting-edge machine learning models for automated cricket shot analysis, enhancing player performance feedback accuracy by 40% using state-of-the-art computer vision and deep learning techniques.Led a high-impact collaborative project with AIIMS Bhopal, creating an AI-powered pre-diagnostic symptom elicitation interview agent that improved patient pre-screening accuracy by 35% and streamlined the healthcare intake process.Secured the prestigious NIDHRI PRAYAS Entrepreneurship Research Grant from iHub-Drishti, IIT Jodhpur, receiving INR 6,00,000 for MVP development and startup establishment. Show less

Developed and deployed end-to-end Computer Vision solutions for Quality Assurance and Quality Control in industrial settingsBuilt robust Flask APIs to facilitate communication between deployed CV solutions and servers, ensuring seamless integrationConfigured and deployed Computer Vision models on Amazon EC2 instances, optimizing for performance and scalabilityCreated user-friendly GUIs for both web-based and RPI-deployed solutions, enhancing user interaction and… Show more Developed and deployed end-to-end Computer Vision solutions for Quality Assurance and Quality Control in industrial settingsBuilt robust Flask APIs to facilitate communication between deployed CV solutions and servers, ensuring seamless integrationConfigured and deployed Computer Vision models on Amazon EC2 instances, optimizing for performance and scalabilityCreated user-friendly GUIs for both web-based and RPI-deployed solutions, enhancing user interaction and controlImplemented full-stack development practices, from backend model deployment to frontend user interfacesCollaborated with cross-functional teams to integrate AI solutions into existing production systemsContributed to improving QA efficiency and product capabilities through innovative AI applications Show less

Developed a priority-based collision avoidance algorithm for UAVs, crucial for applications in military, medical, food delivery, and agriculture.Implemented an artificial potential field approach with varying field strength based on priorityCreated a sophisticated testing framework in ROS and Gazebo using the Hector quadrotor model.Achieved 24% faster completion times for highest priority UAVs compared to lowest priority and 87% improvement over baseline… Show more Developed a priority-based collision avoidance algorithm for UAVs, crucial for applications in military, medical, food delivery, and agriculture.Implemented an artificial potential field approach with varying field strength based on priorityCreated a sophisticated testing framework in ROS and Gazebo using the Hector quadrotor model.Achieved 24% faster completion times for highest priority UAVs compared to lowest priority and 87% improvement over baseline algorithms(RIPNA).Contributed to advancing decentralized collision avoidance strategies for increasingly dense airspaces. Show less

Spearheaded drive-by sensing research to optimize Bangalore's public transit system for urban data collectionAnalyzed geospatial big data from BMTC, developing efficient queries to formulate drive-by sensing as a linear integer programming problemUtilized Geopandas for advanced data visualization, providing crucial insights for urban planning decisionsOptimized sensor deployment strategies balancing budget constraints, sensing frequency, and spatial… Show more Spearheaded drive-by sensing research to optimize Bangalore's public transit system for urban data collectionAnalyzed geospatial big data from BMTC, developing efficient queries to formulate drive-by sensing as a linear integer programming problemUtilized Geopandas for advanced data visualization, providing crucial insights for urban planning decisionsOptimized sensor deployment strategies balancing budget constraints, sensing frequency, and spatial coverageContributed to innovative solutions for real-time monitoring of city-level parameters including pollution, traffic, and temperature Show less

Population density estimation is an expensive task when done with computer vision. Wifi signals emanating from smartphones offer a non-intrusive manner to get approximations of density estimation. Project involved an end-to-end solution for density estimation that is low cost, portable and accurate. Built an Internet of Things (IoT) framework involving a portable micro controller (Raspberry Pi), sensors (Wifi Cards), cloud database (ThingSpeak) and mobile app (MIT app inventor) to… Show more Population density estimation is an expensive task when done with computer vision. Wifi signals emanating from smartphones offer a non-intrusive manner to get approximations of density estimation. Project involved an end-to-end solution for density estimation that is low cost, portable and accurate. Built an Internet of Things (IoT) framework involving a portable micro controller (Raspberry Pi), sensors (Wifi Cards), cloud database (ThingSpeak) and mobile app (MIT app inventor) to solve the task. Used normalisation techniques to remove unwanted signals and MAC address hashing to preserve anonymity. Created a rolling time window algorithm that used the queue data structure to maintain traces of active signals. The app was deployed in IISERB’s dining hall and is accurate within ±20% of ground truth. Show less

Conducted comparative analysis of machine learning models for malicious URL detectionImplemented and evaluated Support Vector Machine (94.5% accuracy), Decision Tree (95.5% accuracy), and Neural Networks (95.9% accuracy)Developed feature extraction techniques for URL lexicographic analysisContributed to cybersecurity research by improving automated threat detection methodologiesGained hands-on experience in applying various ML algorithms to real-world security… Show more Conducted comparative analysis of machine learning models for malicious URL detectionImplemented and evaluated Support Vector Machine (94.5% accuracy), Decision Tree (95.5% accuracy), and Neural Networks (95.9% accuracy)Developed feature extraction techniques for URL lexicographic analysisContributed to cybersecurity research by improving automated threat detection methodologiesGained hands-on experience in applying various ML algorithms to real-world security challenges Show less