About Me
Data Scientist with 4+ years of overall IT experience, including 3+ years specializing in machine learning, data analysis, and AI.Experienced in building classification and regression models, developing RAG pipelines with LLMs, and deploying production MLsolutions on AWS. Strong foundation in EDA, feature engineering, model evaluation, and data visualization, with hands-on experience in end-to-end ML productionization using Docker, CI/CD, and cloud platforms
Experience
Data Scientist
Optimalex · Full-time
- Built and evaluated classification and regression models for insurance claims prediction using linear, tree-based, and boosting algorithms.
- Developed models for liability, indemnity, expense, and claim duration, with preprocessing and feature engineering.
- Used ROC-AUC, F1, RMSE, MAE, and R² for model evaluation, with Optuna for tuning and SHAP for interpretability.
- Built a RAG pipeline to extract 26 structured fields from legal claim PDFs using LangChain, Qdrant, BM25, and Gemini.
- Implemented hybrid retrieval and structured LLM extraction with Pydantic and field-level provenance.
Data Scientist
72 DRAGONS, LLC · Full-time
- Developed a real-time monitoring dashboard using Streamlit, reducing manual effort.
- Automated server and social media data collection via cron-scheduled Python scripts.
- Performed EDA on social media metrics to optimize campaign timing and content targeting, contributing to a 15% increase in engagement.
- Built web scrapers using httpx and selectolax to extract target profiles for sales outreach, increasing lead volume by 25%.
- Designed an alerting system with email automation, cutting incident response time by 40%.
- Stored API/server responses in MySQL using Python with structured logging for timeseries monitoring and alerts.
Test Engineer
L&T Technology Services · Full-time
- Automated testing of I2C, SPI, PWM, and ADC peripherals using Python and Pytest, improving test coverage and reducing manual effort.
- Designed and executed test cases based on Jira user stories; performed both manual and automated validation to ensure feature quality.
- Collaborated with cross-functional teams in an Agile Scrum environment, using Git/GitHub for version control and continuous integration.
- Gained hands-on experience with Selenium for web test automation, enhancing end-to-end testing capabilities.
Quality Engineer
Geometry Engineering Innovations · Full-time
- Automated quality inspection reports to improve efficiency and reduce manual errors.
- Developed Excel dashboards to track raw material inventory, ensuring availability and minimizing delays.
- Designed performance dashboards for various departments, analyzing data on a monthly and quarterly basis to identify trends and areas for improvement.
- Analyzed Non-Conformance (NC) reports to identify quality issues and implement corrective actions, enhancing product reliability.
- Conducted raw material verification and inspection, including cutting sample pieces for testing, to ensure compliance with specifications before production.
Education
M.Tech in Data Science and Engineering
BITS Pilani
Work Integrated Learning Programme • 2023 - 2025
Focusing on machine learning algorithms and statistical modeling. The program provides strong foundations in data science with practical industry applications.
Featured Projects
🎵 Hybrid Spotify Recommendation System
A production-grade hybrid recommender system combining content-based and collaborative filtering. Designed for scalability, performance, and automated CI/CD deployment on AWS.
- 🧠 Content-based filtering using cosine similarity on song metadata
- 🤝 Collaborative filtering on a sparse matrix of 9.7M users × 34K songs (330M+ interactions)
- 🎯 Hybrid score: Weighted average of both approaches for better personalization
- ⚙️ ML Workflow: Versioned pipelines with DVC & Git, artifact storage in AWS S3
- 🚀 CI/CD: Docker → GitHub Actions → AWS ECR → CodeDeploy (Blue/Green deployment)
- 🎧 Example: Recommends similar tracks to “Love Story” by Taylor Swift
🏠 Bangalore Apartment Price Analyzer
A production-grade regression system for predicting Bangalore apartment prices based on property features. Designed for data-backed real estate decisions and built with a modern deployment stack.
- 📊 Market Snapshot Dashboard: Live stats like average price, price/sqft, and property listings.
- 🔮 Price Predictor: ML-powered tool to estimate apartment prices based on zone, area, BHK, and more.
- 📈 Analysis Dashboard: Visualize price trends, filter by zone/BHK/type, and view property distributions.
- 🏡 Apartment Recommender: Suggests similar apartments using location, price, and features (cosine similarity).
- 🧪 End-to-end MLOps: Model training, DVC pipeline, Git versioning, and FastAPI + Streamlit deployment.
- 🔎 Web Scraping: Custom scraper using HTTPX and Selectolax for dynamic apartment data collection.
🚚 Delivery Time Prediction App
A full-stack ML application that predicts food delivery times using real-world features and historical data. Built with versioned data pipelines, MLflow experiment tracking, and automated CI/CD deployment to AWS.
- 🧪 ML Pipeline: Data cleaning, feature engineering, training, and model selection via Optuna
- 📁 ML Workflow: DVC for data versioning, tracked via Git and stored on AWS S3
- 📋 Experiment Tracking: MLflow + Dagshub for experiment logging and model registry
- 🚀 CI/CD: GitHub Actions → Docker → AWS ECR → EC2/CodeDeploy with rolling updates
- 🔍 API Endpoint: Deployed via FastAPI with a live Swagger UI
- 🗺️ Example: Predicts ETA for an order from Indore restaurant to urban delivery location with high traffic
💬 YouTube Comment Sentiment Analyzer
A full-stack NLP application that analyzes YouTube comments in real-time using a custom Chrome extension, backed by a machine learning model served with Flask. Ideal for creators to assess public sentiment quickly and visually.
- 🔌 Chrome Extension: Extracts YouTube comments on-page and sends data to the backend API
- 🧠 Sentiment Model: Logistic Regression trained on preprocessed YouTube data with TF-IDF features
- 🧪 Pipeline: Tokenization, stopword removal, n-grams, undersampling, and hyperparameter tuning
- 📦 Backend: Flask-based REST API for serving predictions in real-time
- 🔬 Model Accuracy: Achieved 87.98% test accuracy on YouTube sentiment classification
- 📊 MLOps: MLflow for experiment tracking, DVC for pipeline versioning
- 🔄 Deployment-ready: Extension and backend work together live in-browser
Technical Skills
Languages
Python, SQL
Data & Databases
Pandas, NumPy, MySQL, ETL Pipelines
ML/AI
Scikit-learn, XGBoost, LightGBM, PyTorch, Optuna, LangChain
MLOps
Git, DVC, MLflow, Docker, CI/CD, AWS
Visualization
Matplotlib, Seaborn, Plotly, Power BI
Web
Flask, Streamlit