Banafsheh Nikbakht

• Built a Time-LLM forecasting model with GPT-2 for industrial time series prediction, covering the full pipeline from data preprocessing to GPU-based training on Databricks. • Applied prompt-as-prefix optimization, improving model performance by 14% over baseline and achieving 2% better accuracy than existing static forecasting algorithms used in production. • Collaborated directly with BASF's data team to deploy forecasting solutions for production planning.

Supported data strategy initiatives by building internal web pages to communicate analytical findings across the organization, making data insights accessible to non-technical stakeholders.

• Analyzed website performance using Google Analytics and identified optimization opportunities that measurably improved user engagement across key conversion funnels. • Built and maintained tracking solutions with Google Tag Manager to monitor conversion events and support data-driven product decisions.