AI Engineer for Developers Associate (2026)

• AI Foundations & Governance: Validated knowledge of AI theory, model usage, safety, privacy, and risk mitigation in generative AI systems.
• LLM Engineering: Applied prompt engineering, chatbot implementation, and structured LLM workflows in Python.
• System Design: Designed and managed AI applications following clean architecture and software engineering best practices.
• Assessment: Successfully passed timed theoretical and practical implementation exams.

Personal Project: University AI Chat Bot (2025)

• University Data Integration: Developed a RAG system specifically for Hochschule Kaiserslautern, processing examination regulations (Klausuren), professor contacts, and internship (Vorpraktikum) data.
• Advanced Retrieval: Utilized ChromaDB and OpenAI Embeddings to enable semantic queries, allowing students to ask natural language questions about university life.
• Geometric Analysis: Implemented t-SNE dimensionality reduction to visualize how university documents are clustered in the vector space, ensuring high retrieval precision.
• Deployment: Launched a live chat interface via Gradio, integrated with LangChain memory to handle follow up questions about campus schedules and contacts.

Tech Stack: Python, LangChain, ChromaDB, GPT 4o mini, Plotly, Scikit Learn, Gradio.

LLM Engineering Applications (2025)

• Built end-to-end LLM systems including RAG pipelines and multi-agent workflows.
• Implemented LoRA fine-tuning and retrieval-based generation techniques.
• Developed production-style applications: brochure generator, support agent, meeting summarizer, Python-to-C++ converter.

Minutes AI (2025)

• Hybrid AI Pipeline: Integrated OpenAI Whisper Medium for high precision speech to text and LLaMA 3.1 8B for structured summarization.
• Advanced Quantization: Implemented 4 bit NormalFloat (NF4) quantization via BitsAndBytes to run large language models efficiently on resource constrained hardware.
• Asynchronous Streaming: Developed a real time streaming engine using Python Threading and Hugging Face TextIteratorStreamer for instant UI feedback.
• Adaptive Hardware Logic: Wrote custom logic to automatically toggle between CUDA (FP16) and CPU (FP32) based on available hardware acceleration.
• Production UI: Built an interactive deployment interface using Gradio, supporting multi format audio uploads and live Markdown rendering.

Embedded AI for Motion Control (2025)

• Hardware Deployment: Deployed Neural Networks on Arduino Nano 33 BLE (TFLM) and Intel MAX 10 FPGA (Manual C implementation).
• Manual Inference Engine: Developed a lightweight C-based inference pipeline for Nios II to bypass TFLite Micro compatibility issues.
• Real-Time Performance: Achieved ultra-low inference latency (~130µs on Arduino / ~250µs on FPGA) for high-frequency control loops.
• Edge AI Workflow: Managed end-to-end pipeline: Data acquisition (WAV), Keras training, TFLite conversion, and Live Serial/UART inference.

Tech Stack: Python, Keras, TensorFlow Lite, C/C++, VHDL/Verilog, Quartus Prime, Scikit-Learn.

Time Series Forecasting (2024)

• Developed and compared Dense, CNN, and RNN architectures for multi channel signal prediction.
• Engineered lagged features and sliding window preprocessing to capture temporal dependencies in sensor data.
• Optimized model performance based on a trade off between inference speed and MSE accuracy.

Tech Stack: TensorFlow, Keras, NumPy, Pandas, Scikit Learn.