Florian Cimander

Confidence Measures for Few-Shot Intent Classification Few-shot intent classification is a challenging task to master since it is highly dependent on the available data. As a result, additional confidence measures (CM) should increase the reliability of the classification process. This thesis aims to provide appropriate CMs based on a prototypical network framework for deciding whether to reject a prediction or take another subsequent action.

• Deployed 10+ machine learning prototypes with NLP, LSTM, Fourier Time Series and general statistical evaluation which improved performances up to 40% • Pioneered machine learning and data science with the goal of developing data-driven solutions that simplify and accelerate the workflow of software developers • Directed the topic “UI Tests with Docker” during innovation week

• Increased model robustness with limited data through calculation of confidence thresholds which reduced False-Positives by 50% • Merged state-of-the-art natural language processing models with German use cases • Reviewed scientific papers in the area of model evaluation scores

• Improved my presentation skills in front of 100+ participants • Supervised students in their statistical education within the framework of the lectures

The lecture is mandatory during the Bachelor of Business Mathematics. It contained descriptive and conclusive statistics, which was strengthened by an introduction to the statistics software R and elementary programming. My task was to hold and supervise the tutorials and to correct the homework.

One of three leading german companies for Jeep customization • Designed a homepage and online shop to improve user interface and experience • Initiated the companies own magazine with 2000+ copies sold • Shaped the appearance of the company with a new brand identity

• Evaluated transaction volumes of private customers through Excel to explore hidden potential of target groups for digital insurance applications under PSD2 • Forecasted the competitiveness of leading insurances based on BaFin data • Analysed and rated the cyber-insurances marked

B2B company for tyres, rims, complete wheels & lubricants

I worked independently for various companies, including EnBW & TargoBank