Dedeepya Yarlagadda

• Built robot perception pipelines for autonomous warehouse navigation using ROS2 - designing and publishing image topic nodes from NVIDIA Isaac Sim to ROS2 for real-time object detection and scene understanding. • Developed a digital twin environment of a warehouse in NVIDIA Isaac Sim, enabling sim-to-real transfer of perception models and accelerating testing without physical hardware. • Automated model packaging and deployment using PyTorch, TensorFlow, and ONNX - converting and optimizing trained models into cross-platform, inference-ready formats for both cloud and edge targets. • Containerized robotic applications and ML deployment pipelines using Docker, ensuring reproducible, environment-agnostic workflows across simulation, development, and production. • Deployed generative AI applications on NVIDIA Jetson Orin Nano Super, running ONNX-optimized models for low-latency edge inference - gaining embedded AI experience at the intersection of robotics and computer vision.

• Designed and developed a full lab website and portal using TypeScript and JavaScript, delivering a responsive, accessible web presence for the research lab that improved visibility of ongoing projects, publications, and team resources. • Architected modular, maintainable front-end components following modern web development best practices, ensuring scalable codebase structure and consistent UI/UX across the portal for both desktop and mobile users. • Collaborated with faculty and researchers to gather requirements and translate complex scientific content into clean, navigable web interfaces, demonstrating cross-functional communication between engineering and research domains. • Applied computational techniques to support biology research workflows bridging software engineering and life sciences to drive data-driven scientific discovery within the lab environment.

• Delivered structured office hours and 1-on-1 tutoring sessions for undergraduate mathematics students—breaking down complex quantitative concepts, strengthening student problem-solving skills, and directly improving course performance and retention. • Designed and developed curriculum-aligned study materials, slides, and practice resources — translating dense mathematical theory into accessible, digestible formats tailored to diverse learning styles across the undergraduate cohort. • Evaluated student assignments with detailed, constructive feedback—identifying recurring knowledge gaps at scale and communicating actionable guidance that supported measurable improvement in student outcomes. • Collaborated closely with faculty to align instructional support with course objectives, demonstrating cross-functional communication, ownership, and the ability to operate effectively within a structured academic environment.

• Fine-tuned CNN architectures on EEG signal data to classify student engagement and cognitive attention levels applying transfer learning techniques to adapt pretrained models to domain-specific neurological signal patterns. • Designed and executed the full signal preprocessing pipeline, including noise filtering, baseline correction, normalization, and temporal segmentation of raw EEG signals, ensuring clean, experiment-ready input data for model training. • Conducted systematic deep learning experiments, iterating on CNN layer configurations, kernel sizes, pooling strategies, and dropout rates to optimize classification accuracy on student cognitive state datasets. • Analyzed and interpreted experimental results using performance metrics and visualized signal feature maps to validate model behavior and support research conclusions.

• Engineered scalable .NET Web APIs with optimized SQL databases on Azure, supporting real-time analytics across cloud-native microservices and reducing backend latency for high-traffic endpoints. • Designed and deployed RESTful services enabling seamless data flow across distributed systems, improving data pipeline reliability and supporting 10+ integrated application modules. • Built AI/ML-integrated backend APIs for predictive analytics workflows, bridging software engineering with data science to deliver production-ready, data-driven features. • Automated CI/CD pipelines to streamline deployment, improving release frequency and system uptime across staging and production environments. • Collaborated in agile cross-functional teams, contributing to architectural decisions and code reviews that improved codebase quality and maintainability.

• Trained a Naive Bayes classification model from scratch using TensorFlow and Python to perform multi-class sentiment analysis, classifying text data into positive, negative, and neutral categories. • Built an end-to-end NLP pipeline covering raw text preprocessing, tokenization, stop-word removal, TF-IDF feature extraction, and model training on real-world text datasets. • Evaluated model performance using precision, recall, F1-score, and confusion matrix analysis, iterating on hyperparameters to improve classification accuracy. • Gained foundational expertise in supervised learning, text classification, and applied NLP, establishing the core ML skill set applied in subsequent roles.

• Built a multilingual AI-powered web application serving farming communities, incorporating multi-language support to maximize accessibility for rural users across regional demographics. • Integrated an AI chatbot enabling real-time conversational query resolution on crop yield pricing, pest identification, and insect attack preparedness, replacing manual information lookup with an intelligent, always-available interface. • Implemented a location-based live weather module using a real-time Weather API, delivering hyperlocal forecasts to support data-driven crop planning and harvest timing decisions. • Developed a government scheme navigator that surfaces and deep-links to official pension and agricultural subsidy portals, reducing the barrier for farmers to discover and access eligible welfare schemes. • Designed and deployed a fully responsive front-end using JavaScript, React, and Node.js, ensuring consistent usability across mobile and desktop devices in low-connectivity rural environments.