Machine Learning Computer Vision

SUMAN MADIPEDDI

********@***.*** 602-***-**** linkedin.com/in/suman-madipeddi github.com

SUMMARY

AI and LLMs Engineer with 2+ years of experience developing AI/ML solutions across NLP and Computer Vision. Proficient in fine- tuning LLMs, optimizing inference, and building scalable AI applications. Integrating AI into production environments and optimizing application performance. Passionate about applying cutting-edge ML research to build high-performance, data-driven systems. EDUCATION

M.S. Robotics and Autonomous Systems Expected May 2025 Arizona State University, Tempe, Arizona 3.73 GPA

• Relevant Coursework: Artificial Intelligence, Applied AI and ML, Reinforcement Learning and Perception in Robotics Bachelor of Technology in Mechanical Engineering (Machine Learning Focused) May 2023 Indian Institute of Information Technology (IIIT) 3.3 GPA TECHNICAL SKILLS

Data Sciences: TensorFlow, PyTorch, sklearn, keras, NumPy, Pandas, Matplotlib, CNN ML and AI: Transformers, LLMs, BERT, GPT, Multimodal (BLIP, CLIP, Flamingo), Computer Vision, NLP, AGI, RAG, PDDL, GAN, RNN Programming and Web: Python, C, C++, Java, CUDA, JavaScript, React Native, Node.js, HTML, CSS, Bootstrap, FastAPI Cloud and MLOps: Amazon Web Services (AWS), GCP, CI/CD, AWS SageMaker, Docker, Kubernetes, Pipeline Automation Tools and Database: Git, GitHub, D3.js, SQL, MongoDB, ROS2, Windows, Linux/ Unix, Data mining Certifications: NVIDIA ODSC Hackathon, AWS Solutions Architect Associate, Prompt Engineering for GenAI, LLMs PROFESSIONAL EXPERIENCE

Minor Chores, Florida, USA: AI and LLMs Engineer Jan 2025 – Present

• Developed a personalized recommendation system and was tasked with integrating an AI-powered chatbot leveraging LLMs and ML algorithms into the website and app to enhance user interactions and support.

• Revamped home page UI and improved user interactions on the website and app, contributing to frontend and backend.

• Tasked with optimizing performance, reducing latency, and enhancing user experience on Android and iOS platforms. Arizona State University, USA: Graduate Research Assistant July 2024 – Present

• Applied Proper Orthogonal Decomposition and neural ODE architectures to reduce data dimensionality. Developed PyTorch version of Mesh-Based Simulation with GNNs and CNNs, optimizing Data Pipeline for seamless integration. 1stop.ai, India: Machine Learning Intern Feb 2022 - Apr 2022

• Developed and fine-tuned Hugging Face BERT models for various NLP tasks, significantly enhancing text classification.

• Applied BERT to sentiment analysis, named entity recognition, text summarization, and translation by hyperparameter tuning.

• Utilized PCA for dimensionality reduction to generate 3D data, enhancing targeted marketing segmentation accuracy. PROJECTS

Object Segmentation on ARMBench (PyTorch, R-CNN, ResNet-50) May 2024

• Performed object segmentation on the ARMBench dataset, containing over 50K images and 450,000 labeled object segments.

• Implemented Mask R-CNN in PyTorch with a ResNet-50 backbone, trained on a subset of 10000 images across 3 subsets: mix- object-tote, zoomed-out-tote-transfer-set, and same-object-transfer-set, for object detection and categorization.

• Achieved mean average precision (mAP) of 0.48, 0.41, and 0.48 for mAP50, and 0.48, 0.06, and 0.38 for mAP75 across 3 subsets. Survey Data Analysis for Customer Assistance System (CAS) Performance (NLP, ML) May 2024

• Analyzed airline customer survey data to derive insights and enable precise decision-making to enhance CAS performance.

• Examined survey segments and 9100 airline customers, employing ML to categorize responses into promoter, passive, detractor, and super-detractor segments while leveraging the latest ML frameworks.

• Leveraged NLP to enhance CAS performance by identifying high-potential segments and comprehensive text analytics. Dynamic Path Planning for Turtlebot Navigation in ROS-Gazebo (Search Algorithms, ROS, Gazebo, RL) Sep 2024

• Tasked with optimizing Turtlebot path planning in dynamic, obstacle-filled environments in Gazebo's Can World and Cafe World.

• Developed and optimized BFS, GBFS, UCS, and A* used (Manhattan & Chebyshev heuristics) for Turtlebot navigation in Gazebo.

• Demonstrated A*'s effectiveness in obstacle-dense environments using data structures and algorithms for reliable pathfinding. Tic-Tac-Toe playing Cobot and Object Detection (OpenCV, Python, Yolov5, aikit_V2/AiKit_280M5) Nov 2023

• Designed an AI-driven Tic-Tac-Toe playing Cobot using Computer Vision, aikit_V2/AiKit_280M5 and Python libraries.

• Developed an interface that utilized ArUco markers to partition the game area into a 3x3 matrix for gameplay detection, achieving high accuracy and decision-making through color and shape analysis, resulting in Cobot winning 3 out of 5 matches.

• Implemented YOLOv5 with OpenCV for image prediction and first 2 class labels, achieving a 0.57 confidence level.

Contact this candidate