Negin Lava

● Built CI/CD and optimized ETL pipelines on GCP Dataproc using PySpark, Shell, and YAML, integrated across Hive, Cassandra, and BigQuery to power feature engineering for backend ML fraud detection models, delivering 2X productivity gains and 5X faster data fetch times. ● Redesigned full-overwrite ETL pipelines to incremental batch processing with load throttling and optimized table partitioning on Cassandra, cutting job runtimes from hours to minutes and improving real-time inference read latency by ~10ms. ● Built an agentic RAG fraud explainer (Eli5 + LLM) connecting feature store and transactional database data, automating resolution of ~500 analyst tickets per year and reducing case investigation time from days to minutes. ● Adapted ETL pipelines and Kafka event handling to process new transaction payload fields, updating XML to JSON transformations and refactoring Cassandra, Hive, and BigQuery schemas to support new data structures by maintaining seamless production order processing and unblocking downstream analytics workflows. ● Orchestrated complex SQL MERGE and aggregation queries on 100GB+ datasets using Airflow and BigQuery, enabling cluster-free analytics pipelines for cross-functional teams of 15+. ● Deployed a nightly batch risk scoring pipeline using PySpark, Hive, BigQuery, and Airflow to compute member fraud risk scores across 21M+ accounts, reducing false positives by 96% and cutting manual review volume by thousands of cases per year. ● Built real-time Looker and Grafana dashboards with alerting for DS and SWE partners, providing transparency into pipeline health, blocking decisions, and fraud scoring logic across teams.

Thesis: A Copula-Based Approach for the Endogeneity Problem • Compared the performance results between ordinary least squares and Gaussian copula methods using R programming language. Studied the robustness of Gaussian copulas method using simulation. • Built an ETL pipeline using Apache spark (sparklyr) in R programming language to filter and aggregate the data from US Department of Education to build 4-dimensional arrays. Reduced the total post-processing efforts by 90%.

• Performed regression analysis and hypothesis tests on clinical data for the department of Pharmaceutics to assist researchers. • Conducted biological data manipulation, normalization, interpretation, and visualization for multiple researches. Used libraries such as dplyr and ggplot2 in the R ecosystem.

Course work: Data Mining- Exploratory Data Analysis- Experimental Design- Sample Design- Interpretable Machine Learning- Financial Analysis- Regression Analysis

Course work: Computational Statistics and Pattern Recognition, Topics in Statistics, Combinatorial Optimization on Surfaces, Numerical Analysis II, Partial Differential Equations, Ordinary Differential Equations, Complex Variables