<CODER> & <AI ENGINEER>

I'm Emma-Luisa Richter. A a programmer and AI engineer from Germany. I write machine learning applications in Python and play around with open-source LLMs.

GM FREN!

I'M EMMA

Yep, that's me, the girl on the left.
I am a German girl studying IT security, but my goal is to work in the field of AI.

My passion is first and foremost programming and Machine Learning, but I also got other hobbies like karate and drawing.

I am currently employed at the university in Mittweida (DE) where I graduated from (B.Sc.). This is a project-based position to develop a chatbot tutoring system for blockchain education. I am mainly focusing on improvements such as optimizing the RAG system or creating ML models for better tool recognition.

PYTHON: ML

Creating machine learning models like a conditional-NER model for tool detection. I use this skill to solve problems in both my personal life and my job.

Scikit-learn, PyTorch, SpaCy, NLTK, Transformers

PYTHON: NLP

Fine-tuning natural langiage processing models for specific tasks like text classification or named entity recognition.

SpaCy, NLTK, Transformers

PYTHON: AI

Building and improving artificial inteligence applications like chatbots with open-source LLMs and integrated RAG-systems. This is the main task in my current job.

LangChain, FastAI, ChromaDB, Qdrant, Transformers, FastAPI, Ollama

HTML / CSS

Building websites like this one. This is more of a hobby.

PHP

Basic PHP knowledge for building WordPress plugins and using PHP with MySQL. I acquired this skill in my current job.

JavaScript

Beginner knowledge for handling simple events and creating responsive design as well as server functionality.

I recently presented at NWK25, a local conference for young researchers from various universities, and it was a cool experience.
I shared my work titled "Evaluation of Prompt Engineering Techniques as a Replacement for Traditional NLP in an Educational Chatbot Assistant for Blockchain Technology." I know, it's a complex title, and the concepts are even more intricate, but here is the short version:

In essence, my research covered whether Prompt Engineering — crafting specific instructions for AI models — could fully replace classic Natural Language Processing (NLP) techniques, like training ML models for classification tasks. I tested this on my university's Blockchain Technology Education Chatbot and with different open-source Datasets. I built all with Python and libraries like SpaCy and Transformers.

The results? Prompt Engineering isn't the holy grail. It's fast, flexible, and definitely shines for tasks like summarization, QA, and sentiment analysis. However, traditional NLP often outperforms in areas like classification and Named Entity Recognition (NER). For example, training a simple logistic regression model can be faster and more robust for detecting malicious user prompts (e.g., attempts to exploit LLMs) compared to relying solely on prompt engineering. In any case, both methods - Prompt Engineering and NLP techniques - can enhace a chatbots capabilities if used correctly and in the right amount.

The full paper is available in german.

COMING SOON