AI Data Preparation and LLM Log Parsing Pipeline (IBM Research)
TextText Generation
I participated in building an LLM automated log parsing pipeline for AI model evaluation. My work involved preparing structured textual log datasets and prompting strategies to enhance log-parsing performance. I benchmarked LLM-based parsing against state-of-the-art tools and collaborated with research scientists for production integration.
• Conducted large-scale preprocessing of 33M+ Spark logs for effective LLM evaluation.
• Designed multiple prompt generation strategies, including KNN retrieval of sample data.
• Evaluated models through ablation studies and log dataset annotation for accuracy comparison.
• Benchmarked results against tools like Drain, Spell, SPINE, LogPPT, and ChatGPT.
I participated in building an LLM automated log parsing pipeline for AI model evaluation. My work involved preparing structured textual log datasets and prompting strategies to enhance log-parsing performance. I benchmarked LLM-based parsing against state-of-the-art tools and collaborated with research scientists for production integration. • Conducted large-scale preprocessing of 33M+ Spark logs for effective LLM evaluation. • Designed multiple prompt generation strategies, including KNN retrieval of sample data. • Evaluated models through ablation studies and log dataset annotation for accuracy comparison. • Benchmarked results against tools like Drain, Spell, SPINE, LogPPT, and ChatGPT.
2024 - 2024