Ashna Arora

I specialize in Machine Learning and AI. My expertise includes developing end-to-end ML and data pipelines, from data cleaning and preparation to training models and deploying them using Docker, Kubernetes, and AWS. My projects have spanned over Natural Language Processing, Generative AI and Retrieval-Augmented Generation systems. I also bring 2 years of software development experience with John Deere where I worked on developing different business features in Python, cloud infrastructure with AWS and also on CI/CD pipelines with GitHub Actions. I’m comfortable working with APIs, writing SQL queries, and using tools like Prometheus, FastAPI, and Kafka.

Highlights

During my master’s at Northeastern University, I worked as a Machine Learning Research Assistant, where I designed and developed a system to detect AI-generated text. I engineered an end-to-end ML pipeline that used linguistic metrics, compression-based features, and model explainability tools like SHAP. I trained multiple models, including XGBoost, SVM, and MLP, achieving an impressive 99.21% ROC-AUC on external datasets. This project sharpened my skills in feature engineering, model optimization, and building robust, production-ready ML systems.

After that, I worked as a Research Scientist on my project Modeling Human Attention with Transformers. The goal was to predict which segments of a video would capture the most viewer engagement, inspired by YouTube’s Most Replayed feature. I created a peak-driven segmentation strategy to align transcript segments with audience behavior, corrected for biases in replay data, and fused local linguistic cues with global video context using a transformer-based architecture. The system achieved strong prediction accuracy across genres like sports, education, and technology, and even worked when replay data was not available — proving its practical, real-world value.

I also have two years experience as a Software Engineer at John Deere where I worked on developing backend features in Python, integrating AWS cloud infrastructure, and improving deployment processes. I led the migration of CI/CD pipelines from Jenkins to GitHub Actions, which significantly sped up deployments and improved reliability. I also built an in-house language translation system for database queries that reduced the database size by 97%, cutting storage costs and improving query performance. These experiences strengthened my skills in large-scale software engineering, automation, and cloud-based deployment.

Beyond Work

Outside of my professional life, I enjoy activities that keep both my mind and body engaged. I love to read and explore new ideas through books, hit the gym to stay active, and experiment in the kitchen by trying new recipes and baking. These hobbies help me stay creative, focused, and balanced.