Dakshina Moorthy Ramesh

• Developed a machine learning model to classify and segment Spotify songs into genres based on features such as tempo, valence, energy, danceability, and acousticness.
• Received hands-on training in both supervised and unsupervised learning algorithms, including Logistic Regression, Decision Trees, Random Forests, K-Means, and SVM.
• Project was mentored by an industry expert from Wipro, providing real-time feedback and guidance throughout the project lifecycle.
• Performed data preprocessing, feature engineering, and applied PCA for dimensionality reduction; achieved a silhouette score of 0.63 for optimal clustering performance.
• Utilized tools and libraries such as Python, Scikit-learn, Pandas, Seaborn, and Matplotlib for model building, data analysis, and result visualization.

Project: AI Research and Article Writing Agent (in collaboration with DeepLearning.AI platform)

• Developed a multi-agent AI system using CrewAI and GPT-4.0-mini to automate research, summarization, and article generation tasks.
• Designed specialized agents (Researcher, Summarizer, Writer) and orchestrated their collaboration in a sequential workflow to produce high-quality, ready-to-publish articles.
• Leveraged GPT-4.0-mini for natural language understanding and generation, ensuring clarity, coherence, and factual accuracy in outputs.
• Gained hands-on experience with agent-based task delegation, prompt engineering, and large language model coordination.
• Conducted testing across multiple topics to validate system performance and identified opportunities for future improvements (e.g., Editor Agent, web UI integration).

Project: Cardiovascular Disease Prediction

• Built and compared multiple ML models (Random Forest, SVM, Logistic Regression, KNN, Decision Tree) to predict cardiovascular disease from patient health data.
• Preprocessed data with feature scaling (StandardScaler) and handled model-specific requirements to ensure accurate and fair comparisons.
• Evaluated model performance using classification metrics including accuracy, precision, recall, F1-score, and ROC-AUC; Random Forest achieved highest accuracy (72.2%) and AUC score (0.77).
• Visualized model outputs and key features using Matplotlib and Seaborn to enhance interpretability and stakeholder communication.
• Identified opportunities for future improvements such as hyperparameter tuning and ensemble methods to boost model performance.

• Collaborated with the Amazon client to source, clean, and analyze customer data using Microsoft Excel, Google Sheets, and Python, ensuring data accuracy and actionable business insights.
• Automated the conversion of Excel files to CSV using Excel macros and Python scripts, streamlining document processing and reducing manual effort by 30%.
• Monitored and analyzed campaign performance via Amazon Connect, supporting continuous optimization through accurate data tracking and reporting.
• Designed and implemented data collection tools such as surveys and polls to gather structured input for analysis and reporting.
• Created data visualizations (charts and graphs) to communicate analytical findings to stakeholders and supported sprint planning meetings with product teams.

• Respond promptly to customer inquiries and issues via email and live chat, using the Zendesk ticketing tool.
• Troubleshoot and resolve technical problems related to Hexnode MDM, including installation, configuration, and usage. Contribute to and maintain a comprehensive knowledge base of troubleshooting articles and FAQs.
• Produce user-friendly analytical reports and documentation for 1,000 users, while managing forums and resolving 250+ customer issues, utilizing data-driven insights to improve customer satisfaction. Provide feedback on product improvements and enhancements.
• Continuously learn and develop effective solutions to provide the best possible support.
• Developed customized data analysis scripts tailored for Mac devices using Bash, and for Windows devices using Batch/PowerShell, enabling remote monitoring and control across 500+ devices.
• Resolved critical client issues, including developing Bash scripts to uninstall applications on macOS, addressing telemetry configuration challenges on Windows 10/11, and troubleshooting ThinkShield version discrepancies in MotoOEM Config.