Built an event-driven data pipeline using AWS to capture changes in DynamoDB and stream them in real-time using EventBridge Pipes, Kinesis, and Firehose. Data was enriched and stored in S3, then cataloged with Glue for downstream analytics via Athena.

Built a real-time event-driven data pipeline using SQS, EventBridge Pipes, and AWS Lambda to ingest, enrich, and transform Airbnb booking stream data. Final data is written to S3 and made queryable through CSV exports.

Built a RAG-based chatbot that summarizes financial documents (10-K, 10-Q) by combining retrieval techniques with LLMs. Integrated multiple models (GPT-3.5, LLaMA 2, Gemma 1.1) and evaluated outputs on metrics like faithfulness, context recall, and answer relevancy. Designed prompt templates, UI components, and comparison experiments to reduce hallucinations and improve factual consistency.

Developed a machine learning pipeline to predict customer churn using Random Forest and XGBoost models. Cleaned and engineered features from historical data, evaluated performance using AUC and F1 scores, and built an interactive dashboard to visualize churn risk across segments.

Created a serverless data pipeline triggered by file uploads in S3. A Lambda function filters delivery events and writes clean JSON to a target S3 bucket. SNS sends notifications on success, and the entire pipeline is deployed via CI/CD using CodeBuild.

A modern ELT pipeline built for an EdTech platform, showcasing transformation and orchestration best practices using Snowflake, dbt Cloud, dbt Core, and Dagster. The project includes modular, documented data modeling, automated workflows, and lineage tracking—supporting CI/CD and both managed and open-source orchestration strategies.

Developed a real-time analytics dashboard for Delhivery's supply chain operations using Snowflake and Tableau. Migrated raw CSVs into a cloud data warehouse, modeled the data using a star schema, and built a UI that visually mirrors Delhivery's official website. Dashboard includes revenue, order flow, defect rates, and shipping cost analytics - fully filterable and exportable.

Designed an end-to-end sales analytics solution using Power BI and Microsoft SQL Server based on a fictional enterprise case. Migrated from static Excel reports to dynamic dashboards built with DirectQuery, enabling real-time tracking of sales KPIs, product performance, and customer insights. The project also modeled a 2021 sales budget and aligned all outputs to stakeholder user stories.

Built a Streamlit-based evaluation tool to benchmark OpenAI models using the GAIA dataset. The app enables users to select test cases, query models, compare results with ground-truth answers, collect feedback, and re-evaluate modified steps. Features include persistent feedback storage, interactive charts for visualizing outcomes, and secure API key handling via environment variables.

Designed a modular dbt pipeline in Snowflake to transform raw e-commerce data into analytics-ready data marts. Implemented a layered architecture (Raw → Staging → Marts), with separate Snowflake databases for each layer. Used source blocks, CTEs, and dbt configs to build scalable models and document the entire data flow using DAGs and lineage graphs.

Built an interactive sales and inventory analytics dashboard in Power BI using Amazon sales data. Applied Power Query for data cleaning, DAX for KPI logic, and designed multi-page dashboards covering orders, products, returns, and regional sales. Includes dynamic filters, custom tooltips, and drill-downs to support data-driven decision-making.

Built an event-driven data pipeline using AWS to capture changes in DynamoDB and stream them in real-time using EventBridge Pipes, Kinesis, and Firehose. Data was enriched and stored in S3, then cataloged with Glue for downstream analytics via Athena.

Built a real-time event-driven data pipeline using SQS, EventBridge Pipes, and AWS Lambda to ingest, enrich, and transform Airbnb booking stream data. Final data is written to S3 and made queryable through CSV exports.

Built a RAG-based chatbot that summarizes financial documents (10-K, 10-Q) by combining retrieval techniques with LLMs. Integrated multiple models (GPT-3.5, LLaMA 2, Gemma 1.1) and evaluated outputs on metrics like faithfulness, context recall, and answer relevancy. Designed prompt templates, UI components, and comparison experiments to reduce hallucinations and improve factual consistency.

Developed a machine learning pipeline to predict customer churn using Random Forest and XGBoost models. Cleaned and engineered features from historical data, evaluated performance using AUC and F1 scores, and built an interactive dashboard to visualize churn risk across segments.