Dhruv Khut

Experience

• 

  IEEE SPIT

  Oct 2021 - Dec 2021

  Frontend Developer Intern

  As a Frontend Developer Intern at IEEE S.P.I.T, I led the creation of a news website for students using React.js and Material-UI, incorporating dynamic PDF generation for adaptive content layouts to improve accessibility. To ensure an engaging and responsive user experience, I leveraged Chakra UI and CSS Grid, which boosted user engagement by 10% and enabled seamless compatibility across mobile devices. These contributions combined to deliver a highly interactive and user-friendly platform.

• 

  Oculus S.P.I.T.

  Dec 2021 - Apr 2022

  Head of Marketing at IPL Auction
• 

  DESCON ENGINEERS

  Feb 2022 - May 2022

  Full Stack Developer

  Engineered and optimized Descon Engineers' website using HTML, CSS, React for frontend development, and Node.js for backend integration, significantly improving performance and scalability, while also leading the creation and integration of 3D models using AutoCAD and Blender to enhance project visualization and client presentations.

• 

  SIMPLECRM Development Test Page

  Aug 2022 - Dec 2022

  Machine Learning Intern

  Developed a sentiment analysis model using TensorFlow that improved customer classification accuracy by 21%, predictive accuracy by 15%, user satisfaction by 12%, and reduced false positives by 20%, while also developing an automated response pipeline for ticket management using Flask and AWS, streamlining customer support and improving service efficiency by 5%.

• 

  Indian Institute of Technology, Patna

  Jan 2023 - Jul 2023

  AIML Research and Development Intern

  We developed the first low-light traffic violation dataset in collaboration with local police, comprising over 1,100 anomaly instances. Using YOLO-v8x, we created a real-time detection pipeline that improved accuracy by 13% over state-of-the-art methods. This system was implemented on the IITP campus, connecting a workstation to a network of eight CCTVs with MAC address mapping for coordinated tracking and an automated email alert system for detected violations. This work was published in IEEE Access.Additionally, we proposed a novel modified UNet-GAN architecture for backlit image enhancement, achieving a peak PSNR value of 25.77 on the BacklitNet dataset. We applied this architecture to enhance number plate recognition in real-world scenarios, with the findings published in IEEE Transactions on Instrumentation and Measurements. Show less

• 

  Bhartiya Vidya Bhavans Sardar Patel Institute of Technology Munshi Nagar Andheri Mumbai

  Aug 2023 - May 2024

  Student Researcher

  I created a hybrid framework combining PCA and ANNs with a custom loss function to enhance log anomaly detection, significantly reducing pseudo-positives and demonstrating effectiveness with both simulated and real-world datasets. I also proposed a load transfer method utilizing drone-carried base stations to optimize power distribution in green micro-grid networks, efficiently managing energy deficits and minimizing base station outages with fewer drone exchanges. Additionally, I developed PotholeGuard, an advanced point cloud-based architecture that improves pothole detection accuracy by addressing point cloud sparsity, enhancing local feature capture, and refining structural insights through adaptive modules and structures. Show less