Top Priority 📊 Data Analyst → Data Scientist | 🤖 AI/ML Enthusiast | 🎯 Next Preferred Role: Power BI Developer

• 💼 Data Analyst with 17 months of experience at TCS
• 📊 Specialized in SAP Analytics Cloud & Business Intelligence
• 🐍 Strong in Python, SQL, Power BI, Tableau & Excel
• 🔍 Passionate about data-driven decision making
• 📈 Experienced in EDA, Data Cleaning, Visualization & Dashboarding
• 🤖 Exploring Machine Learning & Generative AI — clustering, RAG pipelines, and agentic workflows**
• 🔬 Built real-world ML projects including customer segmentation and AI-powered document Q&A systems**
• 📚 Continuously upskilling from Data Analytics into Data Science through hands-on projects**
• 🎯 Goal: To evolve from Data Analyst to Data Scientist — combining BI expertise with Machine Learning & AI to solve real-world problems

• K-Means Clustering & Customer Segmentation
• Retrieval-Augmented Generation (RAG)
• Embeddings & Vector Databases
• Exploratory Data Analysis (EDA) for ML
• Agentic AI & Autonomous Workflows

• MySQL, MySQL Workbench
• SAP Analytics Cloud
• Power Query, DAX
• Git & GitHub
• • Scikit-learn (Machine Learning models — clustering, regression, classification)
• FAISS (Vector similarity search / vector database)
• Hugging Face (Pretrained embeddings & transformer models)
• Groq API (Fast LLM inference — Llama models)
• Google Colab & Jupyter Notebook (already in your list, but worth highlighting here too)
• Seaborn / Matplotlib (Statistical data visualization)

Assistant System Engineer
📅 May 2024 – Oct 2025

• Developed 5+ dashboards using SAP Analytics Cloud
• Built data models & KPI reports for enterprise analytics
• Worked on Ledvance Digital Future (LDF) Project
• Ensured data accuracy, integrity & real-time reporting
• Collaborated with cross-functional teams for analytics solutions

• Built an end-to-end ML pipeline on 20,000 employee records to predict attrition, achieving [X]% F1-score and [X] ROC-AUC using Random Forest, Decision Tree, and Logistic Regression
• Identified the top 5 drivers of attrition (e.g. overtime, income, tenure) through feature importance analysis, cutting through 29 raw features to the ones that actually matter
• Segmented the workforce into [X] clusters using K-Means, uncovering a high-risk group with a [X]% attrition rate — [X]x higher than the company average of [X]%

🔗 https://github.com/swaroop456/employee-attrition-analytics

• Built a binary classification ANN (TensorFlow/Keras) predicting e-commerce purchase intent from session behavior data, achieving 95% test accuracy, 100% precision, and 87.5% recall
• Engineered a preprocessing pipeline handling missing values, categorical encoding, and feature scaling on an 8-feature dataset with a 60/40 class split
• Compared SGD vs Adam optimizers on identical architecture, with Adam converging in 96 epochs vs SGD's 100, using early stopping to prevent overfitting
• Designed a 3-layer neural network (16→8→1 neurons) with ReLU/sigmoid activations, achieving an F1-score of 0.933 on unseen test data

🔗 https://github.com/swaroop456/customer-purchase-prediction-ann

• Built a CNN (TensorFlow/Keras) that classifies clothing images into 10 categories, achieving 90.85% test accuracy on the Fashion-MNIST dataset (70,000 images)
• Reduced overfitting through Dropout (0.5) and real-time data augmentation, keeping the train-validation accuracy gap under 1% across 28 training epochs
• Designed a 3-block convolutional architecture (241K parameters) with early stopping, cutting unnecessary training time by ~7% (28 of 30 max epochs)
• Diagnosed model weaknesses via confusion matrix analysis, identifying a 15%+ accuracy gap on visually similar classes (Shirt vs. T-shirt/Pullover/Coat)

🔗 https://github.com/swaroop456/fashion-mnist-cnn-classifier

• Built an end-to-end NLP sentiment classification pipeline on 100 movie reviews, engineering a full text-preprocessing workflow (HTML stripping, lemmatization, negation-aware stop-word removal) that fed two competing model architectures.
• Implemented and benchmarked TF-IDF + Logistic Regression against a Keras word-embedding neural network, achieving 100% accuracy, precision, recall, and F1-score with the TF-IDF approach on the held-out test set.
• Diagnosed and explained a 30-point accuracy gap (100% vs. 70%) between the two models by tracing it to training-data volume, demonstrating applied understanding of when classical ML outperforms deep learning on small datasets.

🔗 https://github.com/swaroop456/nlp-sentiment-tfidf-vs-embeddings

• Built an autonomous research agent in Python that processes a batch of 5+ topics end-to-end (search → summarize → save), generating structured Markdown reports with zero manual intervention per topic
• Designed a modular 3-tool pipeline (search, summarize, save) orchestrated by a single agent function, with a pluggable LLM layer supporting both offline extractive summarization and OpenAI-based abstractive summarization
• Implemented 3 layers of guardrails — input sanitization, keyword-based content filtering, and automatic retry logic (up to 2 retries per step) — to handle malformed input, unsafe topics, and transient failures without crashing

🔗 https://github.com/swaroop456/research-agent-ai

• Analyzed 541,909 transaction records from a UK-based Online Retail Store (Dec 2010 – Dec 2011) using Python
• Engineered RFM features (Recency, Frequency, Monetary) from raw transaction data, aggregating 4,338 unique customers
• Removed outliers using the IQR method, reducing dataset to 3,710 clean customer records before clustering
• Determined optimal K=3 clusters using Elbow Method and validated with Silhouette Score of 0.4600

🔗 https://github.com/swaroop456/Customer-Segmentation-ML-Project

• Analyzed 1,13,698 Amazon order records spanning across India using Python, Excel, MySQL, and Power BI.
• Built a 5-page interactive Power BI dashboard to track KPIs, revenue trends, customer behavior, and fulfillment performance.
• Identified INR 75M+ total revenue with an average order value of INR 663.
• Discovered that Kurta & Set categories contributed 77% of total orders, making them the highest-performing product categories.
• Found that promotional campaigns generated 2.5x higher revenue compared to non-promoted orders.
• Revealed that Expedited Shipping contributed nearly 75% of total revenue, showing strong customer preference for faster delivery.

🔗 https://github.com/swaroop456/Amazon-Orders-Data-Analytics-Project

• Cleaned & analyzed 10,000+ rows using Pandas
• Performed EDA & correlation analysis
• Built visualizations using Matplotlib

🔗 https://github.com/swaroop456/python-pandas-mini-project

• Wrote complex SQL queries (JOIN, GROUP BY, Aggregations)
• Performed structured database analysis

🔗 https://github.com/swaroop456/MYSQL-Classic-models-DB-Analysis

• Worked on 12,000+ rows dataset
• Used 17+ advanced Excel formulas
• Built interactive dashboards & pivot reports

🔗 https://github.com/swaroop456/Advanced-Excel-Retail-Sales-Analysis

• Created 4-page interactive dashboard
• Developed 7 DAX measures
• Identified key insights like 16% attrition rate

🔗 https://github.com/swaroop456/Power-BI-HR-Analytics-Mini-Project

• Analyzed 31,000+ records
• Built interactive Tableau dashboard & story
• Derived insights on sales trends & customer behavior

🔗 https://github.com/swaroop456/Tableau-Store-sales-dashboard

• Built 4-page interactive Power BI dashboard on Microsoft Financial Sample dataset
• Developed 6 DAX measures including Total Sales, Total Profit, Profit Margin %
• Created calculated column for row-level Profit Margin analysis
• Analyzed 700 rows across 5 countries, 6 products & 5 segments

🔗 https://github.com/swaroop456/Power-BI-Financial-Sales-Dashboard

🎓 B.Tech – Mechanical Engineering
Avanthi Institute of Engineering and Technology
📊 CGPA: 7.18

• 📘 SAP Analytics Cloud – Record of Achievement (SAP) Credly

• 🔗 LinkedIn: https://linkedin.com/in/swaroopkumarvathada
• 💻 GitHub: https://github.com/swaroop456
• 📧 Email: swaroop.vathada@gmail.com

🔍 I get genuinely excited when messy data turns into a clean dashboard — and lately I've been just as excited teaching machines to spot the patterns themselves (RAG pipelines, clustering models, you name it) 🤖📊 📊🚀