Associate Data Scientist at BigTapp (2025-11 – Present)
- Led development of the Data Mapping Accelerator (DMA), an 8-stage agentic AI pipeline (classification → concept extraction → LLM agent matching → verification/QA) built with LangGraph multi-agent orchestration, integrating Open Egeria for metadata governance and Kafka for real-time CDC.
- Designed a multi-signal entity/attribute matching model (weighted feature engineering: name similarity, taxonomy synonym matching, semantic embeddings, structural context) and benchmarked Azure OpenAI embeddings (Weaviate) vs. GPT-4 reasoning, reducing manual mapping effort by 70% and cycle time by 60%.
- Automated data profiling and data-quality rule generation; owned a 3NF PostgreSQL schema redesign for mapping candidates and profiling runs, with per-stage LLM cost tracking (MLflow) to guide model/prompt iteration.
- Extended DMA to auto-generate ETL scripts from mapped schemas using LLMs with automated push to Databricks, boosting Data Engineering productivity by 10% and accelerating project delivery by 3x; deployed via Jenkins CI/CD.
Key Technologies: Python, LangGraph, Azure OpenAI, Weaviate, PostgreSQL, FastAPI, Jenkins, MLflow, Kafka, Databricks
Data Science Intern at BigTapp (2025-05 – 2025-11)
- Implemented MDM Golden Record generation via probabilistic matching and deduplication, improving entity resolution accuracy by 35%.
- Developed Data Engineering Copilot – a GenAI assistant generating SQL transformations and automating ETL pipelines, achieving 75% automation of repetitive data integration tasks.
- Enhanced Credit Risk Monitoring with predictive analytics and anomaly detection; deployed microservices via FastAPI, Docker, and Nginx with JWT-based authentication and React frontend supporting 500+ columns and 100+ tables.
Key Technologies: Python, FastAPI, React, PostgreSQL, Docker, LangChain, OpenAI API, Pydantic, Nginx, JWT
