I am an AI / Machine Learning practitioner based in Passau, Germany, with hands-on experience across Natural Language Processing (NLP), Deep Learning, Computer Vision, and applied LLM systems.
My work includes two IEEE conference publications (ICET 2024 · ICIT 2024), surveillance & UAV object-detection pipelines (YOLO, R-CNN, CNN-LSTM), transformer-based knowledge extraction, and deployable prototypes — from Streamlit inference apps to Flask backends and Scrapy data pipelines.
I am actively seeking Werkstudent positions as an AI/ML Engineer in the German tech market.
| Area | Competencies |
|---|---|
| Machine Learning | Supervised learning, model training & evaluation, classical ML (Naive Bayes, Decision Trees) |
| Deep Learning | CNNs, LSTM, BiLSTM, GANs, transfer learning, fine-tuning |
| NLP / NLU | Text classification, extractive summarization, NER, Knowledge Graphs, financial text structuring |
| LLMs & GenAI | BERT / RoBERTa, OpenAI API, prompt-based generation, hybrid retrieval + generation chatbots |
| Generative AI | LoRA fine-tuning (visuals), ComfyUI workflows, Stable Diffusion, GANs, REST API inference pipelines |
| Computer Vision | Object detection (YOLO, YOLOv5, R-CNN), video anomaly detection, medical imaging (CNN), surveillance AI |
| Data Engineering | Web scraping (Scrapy), dataset construction, preprocessing pipelines |
| Research | Custom dataset design, ROUGE evaluation, reproducible experiments, technical documentation |
Python · PyTorch · TensorFlow · Deep Learning · Machine Learning · Generative AI · LoRA · ComfyUI · NLP · Computer Vision · Transformers · BERT · Hugging Face · LLM · RAG · REST API · FastAPI · Streamlit · YOLO · CNN · LSTM · GAN · scikit-learn · pandas · OpenCV · Jupyter
Languages Python · C++ · HTML/CSS
ML / DL PyTorch · TensorFlow · scikit-learn · Ultralytics YOLO · OpenCV
NLP Transformers · BERT · RoBERTa · NLTK · spaCy
LLM / APIs OpenAI GPT-3.5 · Hugging Face ecosystem
Generative AI LoRA fine-tuning · ComfyUI · Stable Diffusion · GANs · visual generation
CV CNN · LSTM · R-CNN · YOLO · YOLO World · Streamlit deployment
Data & Tools pandas · NumPy · Matplotlib · Scrapy · Jupyter · Roboflow
Web / Backend Flask · FastAPI · PostgreSQL · REST APIs · Google Sheets integration
Dev Practices Git · reproducible notebooks · structured project layouts
| Project | Description | Stack |
|---|---|---|
| urdu-text-summarization | IEEE ICET 2024 — Extractive Urdu summarization with custom UCES-v1 dataset; BiLSTM vs. customized BERT; ROUGE evaluation | PyTorch, BERT, BiLSTM, NLP |
| VigilantAI | IEEE ICIT 2024 — Real-time anomalous activity detection (robberies, weapons) in CCTV streams with custom dataset & YOLOv5 | YOLOv5, Ultralytics, Python, CV |
| Knowledge-Graph-Construction-for-Pollution-Regulation | Entity & relation extraction from regulatory PDFs; Knowledge Graph construction with fine-tuned RoBERTa | PyTorch, Transformers, spaCy, NER |
| Transforming-Unstructured-Financial-Text-into-Comprehensive-Tables | NLP pipeline to structure unstructured financial text into tabular data | Python, NLP, Jupyter |
| AIchatbot | Hybrid LLM chatbot — predefined knowledge base + OpenAI GPT-3.5 for open-domain queries | OpenAI API, Python, Jupyter |
| Project | Description | Stack |
|---|---|---|
| Anomalies-Detection-Using-CNN-LSTM | Video anomaly detection with CNN-LSTM on UCSD / UCF-Crime / UBnormal; federated learning + PSO fusion | TensorFlow, CNN-LSTM, Federated Learning |
| UAVYOLO | YOLO training & evaluation pipeline for aerial / UAV object detection | PyTorch, Ultralytics, OpenCV |
| UAV | HMAY-TSF — hybrid multi-scale adaptive YOLO with temporal-spatial fusion for aerial detection | PyTorch, YOLO, Ultralytics |
| Yolo-Deployment | Real-time YOLO World webcam inference deployed via Streamlit + OpenCV | Ultralytics, Streamlit, OpenCV |
| RCNN-for-object-detection | R-CNN-based cat/dog detection with training notebooks | PyTorch, CV |
| CNN_model_for_Bone_Fracture_detection | CNN for medical bone fracture classification | Jupyter, CNN, Medical Imaging |
| ActiveLife | AI-powered motion controller — camera-based pose/actions for endless-runner games | Computer Vision, HCI |
| Project | Description | Stack |
|---|---|---|
| ColorizingSketches | Sketch-to-color generation using GANs (Pokémon dataset) | GAN, Deep Learning, Jupyter |
| Project | Description | Stack |
|---|---|---|
| Ml_models_to_Predict_Spam_Email | Email spam/ham classification (Gaussian NB, Multinomial NB, J48) | scikit-learn, Jupyter |
| Hybrid-Approach-for-Skin-types-Classification | Hybrid model for dermatological skin-type classification | ML, Jupyter |
| AI-Algorithms | Search & optimization algorithms — BFS, DFS, UCS, A*, genetic algorithms, N-Queens, neural networks | Python, Jupyter |
| Scraping-Degree-Programs | Scrapy crawler for structured master's program data (MastersPortal) | Python, Scrapy, JSON ETL |
| Data-Pack | Data processing & packaging utilities for ML workflows | Python |
| Project | Description | Stack |
|---|---|---|
| WallStreets---Real-Estate-Managing-App | Real estate management platform — role-based dashboards, listings, invoices, payroll | Flask, PostgreSQL, SQLAlchemy |
| FitnessBase | Full gym management system — members, attendance, POS, payments via Flask + Google Sheets | Flask, Python, REST |
| gikeats.github.io | E-commerce food delivery frontend — homepage, cart, checkout, auth pages | HTML, CSS |
| Primary Language | Repositories | Typical Use |
|---|---|---|
| Jupyter Notebook | 10 | ML experiments, model training, evaluation |
| Python | 8 | NLP pipelines, CV, deployment, Scrapy, Flask APIs |
| HTML / CSS | 2 | Web apps & e-commerce frontends |
| C++ / TeX | 2 | Systems programming & research documentation |
Portfolio snapshot across 22 public project repositories — grouped by AI/ML domain and technology layer.
 YOLO · CNN-LSTM · R-CNN · Surveillance |
 BERT · RoBERTa · Knowledge Graphs |
 scikit-learn · AI algorithms · evaluation |
 Flask · PostgreSQL · Google Sheets |
 Scrapy · JSON ETL · pandas |
 GANs · LoRA · ComfyUI · REST APIs |
| Domain | Share | Repos | Focus |
|---|---|---|---|
| 👁️ Computer Vision |  |
8 | YOLO, CNN-LSTM, R-CNN, surveillance, UAV |
| 🧠 NLP / LLM |  |
4 | BERT, RoBERTa, Knowledge Graphs, LLM chatbot |
| 🌐 Web / Backend |  |
3 | Flask, PostgreSQL, gym & real-estate apps |
| 📊 ML / Data Science |  |
3 | scikit-learn, hybrid classifiers, search algorithms |
| 🗄️ Data Engineering |  |
2 | Scrapy, JSON ETL, data packaging |
| 🎨 Generative AI |  |
1 | GANs, LoRA, ComfyUI pipelines |
Based on 22 public project repositories (excluding this profile repo).
| Stack Layer | Technologies | Example Projects |
|---|---|---|
| Deep Learning | PyTorch, TensorFlow, BERT, BiLSTM, CNN, LSTM | urdu-text-summarization, Anomalies-Detection-Using-CNN-LSTM, CNN_model_for_Bone_Fracture_detection |
| Generative AI | LoRA fine-tuning, ComfyUI pipelines, Stable Diffusion, GANs, REST APIs | ColorizingSketches, Yolo-Deployment |
| NLP / LLM | Transformers, RoBERTa, OpenAI API, spaCy, prompt engineering | Knowledge-Graph-Construction-for-Pollution-Regulation, AIchatbot, Transforming-Unstructured-Financial-Text-into-Comprehensive-Tables |
| Computer Vision | YOLO, YOLOv5, Ultralytics, R-CNN, OpenCV, Streamlit | VigilantAI, UAVYOLO, UAV, Yolo-Deployment, RCNN-for-object-detection |
| Classical ML | scikit-learn, pandas, NumPy | Ml_models_to_Predict_Spam_Email, Hybrid-Approach-for-Skin-types-Classification, AI-Algorithms |
| Data Engineering | Scrapy, JSON ETL, pandas | Scraping-Degree-Programs, Data-Pack |
| Web / APIs | Flask, FastAPI, PostgreSQL, HTML/CSS | WallStreets---Real-Estate-Managing-App, FitnessBase, gikeats.github.io |
💡 Open to collaboration on NLP, LLM applications, generative AI (LoRA / ComfyUI), and computer-vision projects.
🇩🇪 Standort: Deutschland · Sprachen: Deutsch & Englisch
Building intelligent systems — from research papers to deployable ML prototypes.