Sairindhri Bhattacharya

• Built the machine learning infrastructure for GenWave’s AI shopping assistant, utilizing TensorFlow.js and MongoDB to deliver product recommendations based on user behavior patterns.
• Integrated Supabase and GPT-4o API for backend support, enabling recommendation generation and interaction with AI models for enhanced user experience and smarter decision-making in GenWave's AI shopping assistant.
• Developed batch processing data pipeline using AWS S3, SQL, Apache Spark, and Apache Airflow to process and analyze user shopping behavior data, enabling efficient machine learning model training and analytics.
• Tested among 30 users and implemented real-time monitoring and model performance tracking, ensuring low-latency, high-availability infrastructure for AI-driven shopping insights.

• Led the development of a real-time stream processing data pipeline for TCS OmniStore, an AI-powered e-commerce platform, utilizing Java, Kafka for data streaming, and PySpark with SQL to process transactional logs, enhancing data accuracy and query performance by 50%
• Implemented payment REST API using Typescript and integrated with third-party merchants Paytm and Master Card, improving transaction processing time by 60%
• Build Web application for EDF Energy Billing System using Microservice, React JS and Node JS, JavaScript increasing website responsive efficiency and customer engagement by 25%
• Developed and optimized an ETL pipeline using Apache Spark and Hadoop to process e-commerce transaction logs, user browsing history, and product catalog data, reducing processing time by 40%
• Utilized CI/CD pipelines, Git, Kubernetes, and Docker to streamline deployment processes and ensure high availability and scalability resulting 20% reduction in deployment time and a 35% increase in system reliability

· Economic Dynamics of Conflict: Modeling U.S. War-GDP Impact

Gathered, and synthesize data from diverse macroeconomic data sources to analyze and model the "Impact of America's Involvement in Global Wars on U.S. GDP." Utilized LSTM, GRU, and 1D CNN for advanced multivariate time series GDP forecasting, applying sliding window techniques and hyperparameter tuning to elucidate the economic implications of global conflicts on the U.S. economy. Link

· Restaurant Radar: Machine-learning web application

Processed over 6 million unstructured Yelp reviews, establishing an AWS data pipeline to an S3 Data Lake and structuring data with Athena using SQL to overcome data challenges. Utilized BERT and fine-tuned the LLM (Llama-2-7B model) in PyTorch for nuanced aspect-based sentiment analysis and topic modeling. Designed a web app to showcase restaurants on Google Maps with sentiment analysis visuals, integrating Flask for the backend and deploying on AWS LightSail with DVC automation for continuous delivery. Link

· Price Pulse Model: Scalable House Price Prediction Model

Designed a scalable pricing model using PyTorch, achieved an RMSE of 0.15 and a cross-validation score of 0.91 with robust MLOps pipelines using MLflow, covering the entire spectrum from data preprocessing to model training, deployment, and regular retraining. Proactively tackled data drift issues to guarantee model robustness and relevance. Link