DURLA LAKSHMIVENKATA NAGA VISWANADHA REDDY User

I’m an AI/ML-focused research intern with a B.Tech in ECE (Specialization in AIML), and I’m drawn to end-to-end problems where deep learning, computer vision, and sensor data come together. I work best when I can translate messy real-world inputs into analysis-ready outputs that teams can actually use.

At the National Institute of Ocean Technology (NIOT) under the Govt. of India, I built a UAV image stitching pipeline using ORB, SIFT, and FLANN feature-matching for large-scale coastal terrain mapping. I also implemented orthophoto generation with illumination correction, and converted 2D orthomosaic outputs into analysis-ready formats used in active NIOT coastal monitoring workflows.

My projects mirror the same mindset: deploy models into practical pipelines. For drone-based image processing and classification, I integrated YOLOv5 real-time object detection on orthomosaic imagery and converted 2D mosaics into 3D spatial representations for volumetric terrain analysis. I’ve also designed a custom CNN + Bi-GRU hybrid model with attention for audio-based classification, achieving 87.4% accuracy using Ranger optimizer and Cosine Decay learning-rate scheduling, and delivering a clear improvement over a baseline CNN.

I care about performance and usability, not just accuracy. In sleep apnea detection, I designed a lightweight deep neural network with TensorFlow for real-time monitoring (over 92% detection accuracy) and optimized model size and inference latency for IoT-enabled wearable deployment. I’m motivated to keep building robust, efficient AI systems that can run reliably in real environments.