Sep 2025 – Jan 2026 · 4 mos · Boston, Massachusetts, United States · Remote

• Building Nidhi AI from the ground up - an intelligent financial companion that transforms how people understand and manage their money through conversational AI.
• Vision: We're creating a platform that unifies expenses, investments, and financial insights into a single intelligent system. Users can ask questions in natural language and get instant, accurate answers about their financial data - no complex queries, no spreadsheets, just conversation.
• Database Architecture & Infrastructure
• Designed production-ready PostgreSQL schema, architecting multi-tenant financial data management that scales
• Integrated account aggregator APIs (Plaid, Setu) to pull real-time transaction data across banking, credit cards, and investments through ETL processes
• Built robust data models supporting hierarchical account structures with automated balance consistency validation
• Containerized entire stack with Docker, orchestrating PostgreSQL, pgAdmin, and Next.js for seamless cloud deployment
• Agentic AI & Natural Language Processing
• Architected a modular Text2SQL system that converts conversational queries into SQL using GPT-5
• Implemented Agentic AI workflow with FastAPI, featuring intelligent query understanding, SQL generation, execution, and natural language response synthesis
• Engineered a comprehensive security layer: SQL injection protection, parameterized queries, user data isolation, and input validation
• Starting from a blank canvas, we're on our way to transform an idea into a working prototype. From database design to AI integration, from Docker orchestration to API development - every line of code brings us closer to making financial intelligence accessible to everyone. We're iterating fast, learning from real user needs, and building something that truly matters. Join us!

Jun 2025 – Aug 2025 · 2 mos · Boston, Massachusetts, United States · Remote

Apr 2025 – Aug 2025 · 4 mos · Boston, Massachusetts, United States

• DEtermined is your comprehensive resource for data engineering excellence. Explore DE Prep for curated interview questions and DE Projects for hands-on project walkthroughs.

Jan 2024 – Dec 2025 · 1 yr 11 mos · United States · Remote

• Designed and implemented a scalable financial data pipeline using Airbyte to automate the extraction, transformation, and loading (ETL) of market data into Amazon S3. This reduced query times by 35%, enabling faster analytics and driving $200K in annual productivity gains.
• Optimized 15 critical data workflows by integrating Dagster with Amazon EMR for distributed processing. This optimization saved 300+ hours annually, improved operational efficiency, and streamlined data processing for large-scale datasets.
• Developed 20+ dbt (data build tool) models with comprehensive documentation, leveraging AWS Glue for automated ETL jobs and Databricks Unity Catalog for centralized governance. This reduced time-to-insight by 25%, enabling cross-functional teams to access data faster and make data-driven decisions.
• Managed Amazon Redshift performance by implementing partitioning strategies and optimizing the AWS Glue Data Catalog. These efforts reduced average query runtimes by 40% and lowered cloud infrastructure costs by 15%, resulting in $50K in annual savings.
• Implemented data quality monitoring using AWS Glue DataBrew, setting up automated checks to identify and resolve anomalies. This process detected 90% of data inconsistencies, ensuring high data accuracy and building stakeholder trust in the reliability of analytics outputs.

Dec 2023 – Feb 2025 · 1 yr 2 mos · Boston, Massachusetts, United States

• 1. Contributed End-to-End Machine Learning projects, PR(#99) and PR(#21). Verify the PRs in the link below.

Feb 2023 – Jan 2024 · 11 mos · Boston, Massachusetts, United States · Remote

• Optimized ETL workflows by orchestrating Apache Airflow and AWS Glue, automating data transformation processes, and reducing pipeline runtimes by 40%. This acceleration enabled real-time analytics, cutting executive decision-making cycles by 25%.
• Reengineered legacy ETL pipelines by migrating Excel-based processes to Amazon RDS, automating data ingestion and transformation. Reduced manual data preparation time by 40% and improved reporting accuracy by eliminating 90% of manual errors in financial datasets.
• Enhanced real-time data streaming using Apache Kafka deployed on Amazon EMR, scaling event-driven architecture to handle 10K+ transactions/hour. Lowered data ingestion latency by 20%, ensuring sub-second updates for mission-critical dashboards.
• Architected a hybrid data warehouse by integrating Snowflake with Amazon Athena, enabling SQL-based querying of historical data stored in Amazon S3. Improved trend analysis efficiency by 35% and reduced ad-hoc analytics costs by 20% through serverless query optimization.
• Evaluated and fine-tuned SQL queries for Amazon Athena, optimizing partitioning strategies and metadata indexing. Boosted dataset processing speeds by 30%, ensuring on-time delivery of enterprise-wide compliance reports.

Sep 2022 – Dec 2022 · 3 mos · Boston, Massachusetts, United States · Remote

• 𝗖𝗣𝗖𝗦 𝗟𝗮𝘄 𝗗𝗮𝘁𝗮𝗯𝗮𝘀𝗲 𝗣𝗮𝗿𝘀𝗲𝗿:
• Designed a system to parse and extract key information from civil complaints filed against police departments.
• Automated the extraction of important legal data using cutting-edge technologies like 𝗢𝗖𝗥, advanced 𝘁𝗲𝘅𝘁 𝗽𝗮𝗿𝘀𝗶𝗻𝗴, and machine learning models, including 𝗚𝗼𝗼𝗴𝗹𝗲 𝗣𝗲𝗴𝗮𝘀𝘂𝘀 for text summarization and 𝗙𝗮𝗰𝗲𝗯𝗼𝗼𝗸 𝗕𝗔𝗥𝗧 𝗠𝗡𝗟𝗜 𝗹𝗮𝗿𝗴𝗲 for natural language inference.
• Streamlined the review process for non-technical users, allowing them to quickly identify critical details within lengthy legal documents, often spanning hundreds of pages.
• Deployed this data extraction pipeline, deployed on the 𝗕𝗨 𝗦𝗵𝗮𝗿𝗲𝗱 𝗖𝗼𝗺𝗽𝘂𝘁𝗲 𝗖𝗹𝘂𝘀𝘁𝗲𝗿, creating a scalable and efficient database-ready solution that eliminates the need for manual review.
• My contributions included advising on the machine learning models' implementation and optimizing the text extraction pipeline to ensure robust and accurate data retrieval.
• This project has real-world implications for improving transparency and accessibility in legal proceedings, making it easier to query and analyze public lawsuits without requiring specialized knowledge or extensive manual effort.
• 𝗗𝗶𝗴𝗶𝘁𝗶𝘇𝗶𝗻𝗴 𝗛𝗶𝘀𝘁𝗼𝗿𝗶𝗰𝗮𝗹 𝗛𝗲𝗿𝗯𝗮𝗿𝗶𝘂𝗺 𝗦𝗽𝗲𝗰𝗶𝗺𝗲𝗻𝘀:
• Digitized handwritten herbarium specimen labels using 𝗢𝗖𝗥 and Deep Learning.
• Developed a 𝘁𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺𝗲𝗿-𝗯𝗮𝘀𝗲𝗱 𝗢𝗖𝗥 model and built a plant classifier to extract and analyze key data from century-old specimens. Contributed by integrating the 𝗖𝗥𝗔𝗙𝗧 object detector and 𝗧𝗲𝘀𝘀𝗲𝗿𝗮𝗰𝘁 𝗢𝗖𝗥 with 𝗧𝗲𝗻𝘀𝗼𝗿𝗙𝗹𝗼𝘄, boosting text recognition efficiency by 𝟯𝟬% and reducing manual data entry time.
• This work supports climate research by making historical biodiversity data accessible and analyzable for climate change studies.

Jan 2022 – May 2022 · 4 mos · Boston, Massachusetts, United States

• Teaching Assistant for the course CS 677 Data Science with Python under Prof. Eugene Pinsky

Jun 2019 – Aug 2020 · 1 yr 2 mos · On-site

• Deployed Mask-RCNN model for document field extraction, increasing accuracy to 97%, reducing errors in financial document processing by 7%, and cutting operational costs by $10K monthly.
• Automated data entry workflows using Docker and Flask, cutting manual effort by 80% and saving 200+ work hours monthly.
• Designed and implemented an MLOps pipeline with GCP services like Cloud Run and Cloud Build, enabling seamless deployment of updates, and enhancing system uptime by 15%.
• Leveraged real-time monitoring tools (Cloud Monitoring and Logging) to proactively identify model performance issues, maintaining 95% precision across deployments.

Feb 2019 – Jun 2019 · 4 mos · On-site

• Performed a collaborative approach with the company's software development team and product manager.
• Digitized over four scanned documents and financial statements using Google OCR to retrieve required fields from documents with an accuracy of 90%.