ML/AI Engineer

Multi-tenant AI Platform (Production)

KIconnect - RWTH IT Center | Full-time | Sep 2024 - Present

Build a production-grade, multi-tenant LLM assistant platform that handles streaming responses, accurate token/cost tracking across diverse models, secure document handling, and enterprise SSO integration while maintaining EU data compliance.

• Unified Tokenization: Integrated tokenization for both text and vision across OpenAI, Llama 3, Gemma 3, Mistral, Qwen, and DeepSeek models using Microsoft.ML.Tokenizers and custom Python bridges via pythonnet
• Real-time Streaming: Implemented streaming chat with SignalR for real-time token-by-token responses with precise cost tracking
• Admin Dashboard: Built comprehensive Vue 3 + Inertia admin UI for model/deployment management, user quotas, and reasoning effort configuration
• Security & Compliance: Integrated Shibboleth SSO, MongoDB Client-Side Field Level Encryption (CSFLE), EU data region compliance
• Reliability: Implemented retry policies with Polly, circuit breakers, and graceful degradation

KIconnect Platform (Internal)

Time Series Generative Models for Traffic Scenario Generation

RWTH Cyber-Physical Mobility Group | Master Thesis | Oct 2023 - Apr 2024

Implement and train time series generative models to synthesize realistic vehicle trajectory data for autonomous vehicle testing. Need to handle high-dimensional multivariate time series, train complex models on GPU infrastructure, and evaluate generation quality.

• Model Implementation: Implemented TimeGAN (GAN-based) and Diffusion-TS (diffusion-based) architectures in PyTorch for multivariate time series generation
• Data Pipeline: Processed IKA real-world driving datasets (inD, rounD, exiD) with sliding window extraction, normalization, and data augmentation
• GPU Training: Trained models on NVIDIA H100 and Quadro RTX 6000 GPUs with hyperparameter optimization, learning rate scheduling, and early stopping
• Model Evaluation: Implemented evaluation pipeline using PCA, t-SNE visualization, and statistical metrics to assess trajectory realism and diversity
• Integration: Built conversion pipeline to XML format for CPM Remote platform, enabling generated scenarios to be used in autonomous vehicle simulation
• Production Deployment: Packaged models for inference, implemented batch generation, and integrated with existing testing infrastructure

GitHub Repository | Thesis PDF

LLM & GNN Research for Medical Image Analysis

RWTH Chair of DBIS | Research Assistant | Jul 2023 - Mar 2024 (9 months)

Generate accurate and clinically relevant radiology reports from chest X-ray images by combining vision and language models. Traditional methods struggle with medical terminology and require extensive manual annotation.

• Developed a pipeline combining BLIP (vision-language model) with LLMs for chest X-ray report generation
• LLM Deployment & Serving: Deployed LLaMA models on GPU cluster using Ollama, built REST API service with FastAPI for internal research access, managed model inference and user requests
• Conducted research on integrating Knowledge Graphs with GNNs to improve medical domain understanding
• Ran large-scale experiments on GPU cluster infrastructure for model training and evaluation, experimented with vLLM for high-throughput inference
• Built end-to-end LLM chat platform: GPU deployment → API service → web interface for research demonstrations

Semantic Kernel with Local LLMs

Personal Project | Apr 2024

Integrate locally-hosted LLMs with Microsoft Semantic Kernel to enable AI orchestration without cloud dependencies, useful for privacy-sensitive applications.

• Developed connectors to integrate local LLM endpoints with Semantic Kernel
• Implemented function calling and semantic memory for local models
• Demonstrated how to build AI agents using on-premise infrastructure

GitHub Repository