Ajay L

About Ajay L

I\'m a Data Engineer with over 6 years of experience building the data infrastructure that…

Experience

1. Data Engineer

   Vanguard

   Aug 2023 — Present · Malvern, PA, US

   Built real time data pipelines using Apache Kafka and Spark Streaming to power Vanguard\'s AI investment chatbot, handling over two million client interactions daily with consistently low latency.• Designed and implemented ELT workflows on AWS Glue and Snowflake to transform client portfolio, transaction history, and behavioral data into formats ready for LLM training and live inference.• Set up vector database infrastructure using Pinecone to enable retrieval augmented generation for the chatbot, improving response accuracy significantly for investment-related queries.• Developed feature engineering pipelines extracting over 50 client behavioral signals including risk tolerance, asset allocation drift, and investment frequency to feed personalization models used by the advisor tool.• Applied PII anonymization and field level data masking across all AI training datasets to meet SEC, FINRA, and CCPA compliance requirements.• Built dbt models in Snowflake to automate expense ratio tracking across 400 plus funds, contributing to reporting on $600 million in projected investor savings for 2025 and 2026.• Designed a star schema data warehouse for fund analytics, reducing the finance team\'s manual reporting effort by approximately 70 percent.• Conducted Spark performance reviews on AWS EMR, resolving shuffle bottlenecks and partition skew issues to improve pipeline throughput by over 30 percent.• Maintained Airflow DAGs with automated alerting and retry logic, keeping production pipeline availability above 99.9 percent.• Led onboarding of data sources from CRM systems, trading platforms, and third party market data providers into a centralized Snowflake platform used across business units.• Partnered with product managers, ML engineers, compliance officers, and business analysts to translate complex investment domain requirements into well-documented, maintainable data pipelines.