Diffusion Model Experts in Germany

• Fine-tuned and evaluated alternative backbone architectures for the AFFGAN handwriting style encoder to better capture writer-specific features, benchmarking against the original baseline architecture and replacing the legacy VGG19 architecture with a custom Transformer-based encoder.
• Extended the system beyond word-level synthesis to line-level handwriting generation, integrating and adapting an existing line-generation architecture to support full-sentence output.
• Deployed the handwriting generation pipeline as a web application, enabling users to select a writing style, input custom text, and instantly generate a corresponding handwriting image.
• Integrated TrOCR into a teacher–student distillation framework to evaluate and improve readability of generated handwriting, improving OCR accuracy by 20% over the baseline.
• Boosted user-study pick-rates by 40%, demonstrating a significant leap in the perceptual quality and realism of the generated handwriting compared to the baseline model.

Tech: PyTorch, Python, GANs, CNN, ViT, TrOCR, Streamlit, CUDA.

• Developed and implemented AI-based perception and multimodal systems for real-world environments
• Built, trained, and evaluated Machine Learning and Deep Learning models using Python, PyTorch, TensorFlow, and OpenCV
• Worked with Vision-Language Models (VLMs), Large Language Models (LLMs), transformer-based architectures, and multimodal AI systems
• Applied LoRA-based fine-tuning techniques and experimented with diffusion models for generative and multimodal AI applications
• Developed multimodal perception pipelines using camera, LiDAR, and sensor data
• Designed end-to-end workflows for data processing, model training, evaluation, benchmarking, and robustness analysis
• Utilized HuggingFace Transformers and modern Deep Learning frameworks for AI experimentation and deployment workflows
• Applied GPU-accelerated computing, CUDA-based processing, ONNX, and TensorRT optimization for efficient inference and large-scale model training
• Collaborated with industry partners including Valeo and REHAU on applied AI and intelligent system projects
• Developed scalable AI architectures and prototype software solutions for automation and perception tasks

• Developed end-to-end reproducible ML pipelines (PyTorch) with data versioning (DVC), experiment tracking (MLflow), automated testing (PyTest), and CI/CD across all training workflows.
• Scaled Vision Transformer and CNN training across NVIDIA A100 GPU clusters (CUDA, DDP, SLURM); applied hyperparameter optimization (W&B Sweeps) to reduce training overhead and identify optimal configurations.
• Developed a real-time 3D human motion generation system (ViT, VQ-VAE, SMPL-X/PIXIE) for personality-conditioned avatar synthesis; achieved state-of-the-art FID = 6.15 and P-FID = 10.31 on the UDIVA benchmark.
• Validated model expressiveness through structured user studies, achieving 86% accuracy in distinguishing extroverted vs. introverted avatar behaviors.
• Optimized inference pipelines by deploying PyTorch models via TensorRT and ONNX Runtime into native C++ code; benchmarked performance.

• Developed a React-based research platform with interactive 3D/AR product visualization workflows.
• Implemented interaction tracking and usage analytics to measure user behavior and feature engagement within a research platform.
• Deployed the research platform and integrated a R Shiny analytical dashboard, enabling researchers to interactively explore meta-analysis results within a unified platform.

• Integration of Post-Quantum Cryptography (PQC) algorithms into high level protocols.
• Security of implementations of Post-Quantum Cryptography algorithms.
• Transition to Post-Quantum public key infrastructures.
• Security evaluations of Post-Quantum Cryptography (PQC) primitives.
• Drone Cybersecurity
• Satellite Cybersecurity
• AI Security

• Independently initiated and deployed automated validation workflows using Python, cutting manual processing by 58% and improving efficiency
• Developed a machine learning model for synthetic defect generation, reducing downtime and production costs; deployed locally and via Databricks and Azure AI Factory
• Utilized a small dataset of image data from the production lines and extended this dataset with training on models like cycleGAN and pix2pix
• Built and optimized the Linux-based development environment for training 3D models; maintained reproducibility via GitHub
• Presented technical insights to cross-functional teams (engineers, QA, project managers), ensuring alignment of ML solutions with operational needs

• Generated synthetic datasets for CXR, mammography, and distal radius fracture detection using GANs and diffusion, creating >50k synthetic images for benchmarking.
• Ensured GDPR-compliant workflows and reproducibility, enabling dataset adoption for internal validation and academic collaboration.
• Project highlighted in MedTank’s internal R&D showcase as a flagship synthetic data initiative.

• Consulting and interim management for portfolio strategy and implementation
• Building transition and monetization portfolio, focus on telecoms
• Business unit transformations (structure, services, roles)
• Executive coaching
• Go-to-market, positioning, and offering logic
• Transferring telecom-specific monetization logics to other industries

• Creation and planning of the content plan
• Editorial planning for 10 channels
• Image shoot (organization, coordination, execution)
• Selection of image, text and video content

• Architected and deployed LLM based AI agents, RAG pipelines, and vector search solutions for decision support across retail banking department.
• Developed and shipped robust AI pipelines with guardrails, error handling, monitoring, and fallback logic ensuring high reliability outcomes and compliance with data privacy.
• Developed and deployed ML models to identify transactional anomalies, improving fraud detection and risk assessment in high-volume datasets for credit risk modelling.
• Built, evaluated and fine-tuned ML models to generate propensity scores for customers used to drive personalized targeting campaigns for credit cards and personal finance/loan products.
• Developed an NLP pipeline using BERT embeddings and spaCy NER for SMS/email analysis and customer query logs.
• Trained machine learning models using Isolation Forest to classify user behaviour and detect anomalies.
• Extracted, cleaned, enriched and feature engineered datasets from different sources to build feature stores that powered ML model training.
• Led development of dashboards using Power BI, Grafana, and Prometheus to monitor model performances, KPI trends, and marketing metrics.
• Built multi-touch attribution models using logistic regression and time-decay weights to evaluate lead quality.
• Developed scalable ETL pipelines from CRM, T24, SAP, and ERP, supporting millions of monthly transactions.
• Integrated testing and CI/CD workflows for robust data pipeline deployment.

• Gathered tire images and their feature descriptions.
• Cleaned dataset of image metadata using pandas.
• Stored image feature embeddings in Chroma vector db.
• Used image augmentations to increase dataset size.
• Used sklearn to create shuffled datasets and imbalanced-learn to balance class sizes in dataset.
• Used PyTorch to train and test different neural networks.
• Validated model using custom accuracy metric based on similarity search in ChromaDB.
• Visualized accuracy predictions using matplotlib.
• Plugged trained model into DreamBooth to train stable diffusion model and generate new images of tires.
• Created custom Docker image in Amazon Elastic Container Registry for machine learning script.

• Designed and implemented a real-time perception pipeline using YOLOv7 on Time-of-Flight (ToF) sensor data, enabling live streaming, inference, and on-frame visualization for passenger detection.
• Fine-tuned and evaluated multiple state-of-the-art monocular depth estimation models for Automatic Passenger Counting (APC), and developed a custom hybrid depth model that improved depth accuracy in challenging scene regions.
• Demonstrated that model-generated depth maps outperform raw sensor depth for APC tasks across several datasets, contributing to measurable reductions in counting error.

• Led a team of junior Research Engineers, providing mentorship, technical guidance, and career development support to foster their growth in deep learning and machine learning engineering.

• Investigating the equivalence between the Tiny Reasoning Model (TRM) and diffusion models for structured reasoning tasks such as Sudoku and maze solving.
• Exploring the reasoning and generative capabilities of diffusion models in symbolic problem-solving environments.