Internship • Machine Learning/NLP
About The Project
Developed large-scale NLP pipelines and agentic LLM workflows to extract anomalies and failure symptoms from millions of service records, and trained predictive models for repeat and 90-day failure recurrences. Delivered components for a dealer-facing dashboard to inform proactive service and parts planning.
Achievements
Built Spark-based agentic framework using LLM prompting for anomaly extraction at scale
Fine-tuned domain-adapted LLaMA-2 for structured field alignment (missing/inconsistent data)
Engineered NLP stack (TF-IDF, KeyBERT, BERTopic) on 2M+ service records
Trained XGBoost models for repeat failures & 90-day recurrence; integrated into dealer forecasting
More Projects
Links
sunny.sunho.park@gmail.com
