Narendra Devarasetty

Staff Data Engineer · San Francisco, CA

• Re-architected Unit Economics platform, consolidating fragmented revenue, cost, and contribution-margin logic into canonical datasets powering finance close, forecasting, and product analytics across 10+ orgs.
• Delivered GPU efficiency initiative in 4 weeks, partnering with Cloud Infrastructure to analyze $1.13M/month GPU spend and identify 35–50% cost reduction potential ($400K–$600K/month) across 3,000+ workloads.
• Served as technical owner for cross-functional data alignment, translating ambiguous business questions into durable, production-grade data models adopted by Finance, Product, and ML teams.

Staff Software Engineer, Data · San Francisco, CA

• Owned backend architecture for AI-driven voice ordering system delivered in 40 days, including order-cart construction, LLM-optimized menu modeling, and inference integration for a high-visibility client demo.
• Built synthetic data generation pipelines enabling continuous fine-tuning of GPT-3.5 and LLaMA models, improving order accuracy and accelerating iteration without reliance on scarce real-world conversations.
• Built LLM-powered data discovery and SQL generation tooling used by 100+ analysts, reducing ad-hoc analysis time by abstracting schema complexity through metadata and semantic search.
• Designed a config-driven metrics platform powering 1,000+ Tier-0 metrics across dashboards, experimentation, and segmentation, eliminating bespoke SQL and standardizing metric logic.
• Built and scaled a next-generation ETL platform (Spark, Snowflake) with CI/CD, testing, observability, SLA enforcement, RBAC, and data quality checks, adopted by 20+ teams and scaled to 2,500+ production workflows.
• Led Data Reliability Engineering (DRE) sustaining 2+ years of SLA compliance for 150+ Tier-0 metrics across 400+ upstream tables, reducing critical dashboard downtime by 3–4 hours.
• Owned quarterly earnings metrics and partnered with Finance, Product, and PwC to support audit and certification requirements.
• Led company-wide data engineering cost efficiency initiative, delivering $2M in savings in H2 2024 and defining a 2-year roadmap for sustainable cost governance.

Lead Data Engineer · San Francisco, CA

• Built User & Team 360 canonical datasets with 400+ attributes, establishing a single source of truth supporting Product and Finance analytics across 700M+ users.
• Created Dropbox Usage DataMart modeling file activity (adds, updates, shares, collaboration) over 1-day, 7-day, and 28-day rolling windows, ingesting and aggregating billions of file events per day at scale.
• Built Finance DataMart for User and Team ARR, integrating dimensional entities to deliver trusted financial reporting.
• Re-engineered large-scale filesystem usage pipelines using Spark, reducing ETL runtimes by 10+ hours.
• Developed reusable Airflow operators for Hive/Spark workloads with built-in validation and structured logging, improving pipeline reliability and alerting.

Data Engineer · New York, NY

• Built one of Goldman Sachs' first Hadoop-based data lake platforms, leveraging HDFS, Spark, Scala, and Pig to modernize large-scale analytics and batch processing.
• Designed automated ETL frameworks using Spark/Scala and Python, significantly improving data processing efficiency and operational reliability.
• Migrated 350+ legacy BI reports to the modern data platform, optimizing batch job runtimes by 2+ hours and reducing maintenance overhead.
• Led data onboarding initiatives to AWS and Snowflake, enabling early cloud adoption and scalable analytics workflows within SecDiv.

Data Engineer · Nashua, NH

• Consolidated 500+ batch jobs into 25+, improving reliability and maintainability of fixed-income data pipelines.
• Built real-time Trade DataMart with 5-minute SLA, implemented Oracle VPD for data security, and led early Hadoop POCs for scalable analytics.