Looking for a job

Aayush User

@aayushuser14

Message

Data Scientist at PACCAR who deployed an NLP pipeline across 15M+ warranty summaries.

United States

Message

At PACCAR, I built and deployed an NLP classification pipeline across 15M+ warranty summaries, cutting manual claim review by one hour per day. I used Python and PySpark to engineer features, and deployed the workflow on AWS EC2 for recurring batch execution.

At Mu Sigma Inc., I built and owned replenishment pipelines combining ARIMA and Prophet forecasting, automating shipment recommendations across 30K+ SKUs and four warehouses. The work contributed to $8M in EBIT savings and helped improve online fill rate from 20% to 40% in Spain, Portugal, and France.

My work also includes production machine learning, experimentation, and analytics. I developed a CatBoost assortment model that reduced split shipments by 5%, and at CausifyAI - University of Maryland, College Park, I built agentic AI workflows and developed causal inference analyses.

Experience

Work history, roles, and key accomplishments

Advanced Machine Learning Teaching Assistant

University of Maryland, College Park

May 2025 - Dec 2025 (7 months)

Mentored 140 students on PyTorch/TensorFlow, transformers, CNNs, RNNs, LSTMs, and Bayesian methods. Developed Bayesian tutorials and Docker-based environments to standardize project execution and evaluation.

PyTorch TensorFlow Transformers CNN RNN LSTM Bayesian Statistics Docker

Data Scientist Intern

PACCAR

May 2025 - Aug 2025 (3 months)

Built and deployed an NLP classification pipeline across 15M+ warranty summaries, cutting manual claim review by 1 hour per day. Engineered TF-IDF and embedding-based features using Python and PySpark; benchmarked models with MLflow, with Random Forest achieving 0.80 weighted F1.

Python PySpark NLP TF IDF Embeddings MLFlow Random Forest SHAP LDA Snowflake Tableau EC2

Data Scientist

Mu Sigma Inc.

Oct 2021 - Jun 2024 (2 years 8 months)

Built and owned end-to-end Databricks, PySpark, and Airflow-based replenishment pipelines combining ARIMA and Prophet forecasting to automate STO shipment recommendations across 30K+ SKUs and 4 warehouses, contributing to $8M in EBIT savings. Improved online fill rate from 20% to 40% over 18 months in Spain, Portugal, and France by forecasting demand and optimizing inventory allocation.