Kujtim Rahmani
Verified Expert in Engineering
Machine Learning Developer
Kujtim is a computer vision and machine learning engineer with over ten years of experience and more than four years of experience in academia. During his tenure, he worked with four startups, where he built products from scratch. Furthermore, Kujtim also worked as a consultant for esteemed clients such as the German automotive industry and Heraeus Group and as a data scientist researcher for Airbus.
Portfolio
Experience
Availability
Preferred Environment
Ubuntu, Windows, Visual Studio Code (VS Code)
The most amazing...
...thing I've developed is the SOTA algorithm for defect detection for airplane parts. I used traditional ML to build SOTA for segmentation in a small dataset.
Work Experience
AI/ML Expert and Consultant
Jack Zerby Consulting, LLC
- Created an algorithm for the intelligent processing of collectible images. Developed several supervised detection and intelligent character processing of collectibles. Handled several templates.
- Helped the team integrate the service in the AWS and contributed to designing the app.
- Fine-tuned several deep learning models to process and better recognize the text in the images, as well as handle the errors made by the recognizer.
Senior Data Scientist | Computer Vision
Pano
- Built computer vision algorithms for fire detection and tracking in forests.
- Developed mathematical modeling of the fire initiation and growth during the night. Used Python, OpenCV, scikit-learn, GCP, and GitHub.
- Created a tracking algorithm for tracking forest fires during the night.
Senior Computer Vision Engineer
Rare Edition
- Built computer vision algorithms for card and coin defect evaluation. Conducted quality assurance for collectibles.
- Communicated with operations (graders), back-end services, and the mobile developer to ensure all computer vision algorithms ran correctly.
- Created an active learning platform for card grading.
Machine Learning Engineer
CollX
- Developed an image search and matching engine for sports like baseball, basketball, and soccer and game cards such as Pokemon and Magic. The users can take a photo of their card, and my algorithm finds the matching card in the system.
- Used contrastive and triplet loss metrics to build an image embedding system.
- Created a hard example mining system for algorithm improvement.
- Built a Faiss indexer to index the card images and integrated feature matching to filter the best results.
Senior Data Scientist | Computer Vision
Ramani
- Built computer vision algorithms for object detection, counting, and tracking in warehouses.
- Conducted 3D reconstruction of the objects based on multiple images.
- Managed junior computer vision engineers and annotators.
Senior Data Scientist | Computer Vision
Smartauger
- Built computer vision algorithms for objects like pipes and optical fiber cables and their detection in electromagnetic wave-generated images from ground-penetrating radars.
- Ran experiments and built the algorithm using simulated and real data.
- Guided the stakeholders on how to acquire the data from real scenarios.
Senior Data Scientist | Computer Vision
Heraeus
- Built computer vision algorithms for object detection and tracking of defects of manufactured parts in X-ray images.
- Optimized the detection and tracking to process more than 17 images per second.
- Collaborated with the clients to ensure the data and annotation were of the best quality.
Machine Learning Engineer
Tolstoy
- Built an OCR automatic parsing of the electricity bills.
- Integrated Google Drive with the Cloud Run and Cloud Scheduler to ensure new data was parsed after being uploaded to Google Drive.
- Handled several electricity bill templates and file formats.
Data Scientist Consultant
Cluster Reply
- Built an object detection and tracking algorithm for counting objects in 3D computed tomography images of vehicle parts. Reduced the time from around 120 minutes to five minutes.
- Created an invoice data extraction algorithm for a private German digital post service.
- Developed NLP algorithms for a research paper, understanding and computing its correlation to each of the UN countries' sustainable development goals.
- Prepared machine learning and data science project proposals for customers.
Data Scientist
Airbus
- Built a pore detection algorithm in 3D CT scans of airplane parts with 99.9% accuracy.
- Created a system design for change detection in airplane cabins using radio frequency signals.
- Guided the material scientist on data annotation and understanding how ML algorithms work.
Research Assistant and Scientist
UniBW
- Developed a semantic scene segmentation algorithm for facade images, which is state of the art for small datasets.
- Created a dataset for detailed facade segmentation, including transom and mullion windows. Analyzed and considered data for creating high-quality 3D building models.
- Built a complete pipeline for facade segmentation from data acquisition up to deployment.
Machine Learning Engineer
Zyncd
- Built a matching algorithm to match user profiles with problem descriptions.
- Improved the accuracy of the matching by more than 50%.
- Developed a ranking algorithm using mathematical functions to normalize profiles.
- Integrated word2vec into the system for profile embedding.
Computer Lab Teacher
Ss. Cyril and Methodius University in Skopje
- Held calculus courses for first and third-year students.
- Conducted discrete mathematics courses for first-year students.
- Taught a course on the basics of IT for first-year students.
Experience
Image-searching Engine for Cards
http://collx.appPore Detection Algorithm in 3D CT Scans
Object Detection in 3D-computed Tomography Images
Each CT image was around 1GB. Since there was no pre-trained model for 3D images, I converted the problem to object detection in each slice of the 3D image and then used tracking and dynamic programming to recreate and delineate the 3D objects.
Semantic Scene Segmentation Algorithm for Façade Images
https://isprs-annals.copernicus.org/articles/IV-2/223/2018/isprs-annals-IV-2-223-2018.pdfInformation Extraction from Collectibles
Education
PhD in Artificial Intelligence
University of the Bundeswehr Munich - Munich, Germany
Master's Degree in Computer Science
Technical University of Munich - Munich, Germany
Bachelor's Degree in Computer Science
Ss. Cyril and Methodius University in Skopje - Skopje, Macedonia
Certifications
Certified Microsoft Trainer
Microsoft
Azure AI Associate
Microsoft
CertNexus Certified Artificial Intelligence Practitioner (CAIP)
CertNexus
CertNexus Certified Artificial Intelligence Practitioner (CAIP)
CertNexus
Microsoft Certified: Azure Data Scientist Associate
Microsoft
Skills
Libraries/APIs
OpenCV, NumPy, TensorFlow, Keras, PyTorch, Scikit-learn, Matplotlib, Pandas, Azure Cognitive Services
Tools
You Only Look Once (YOLO), Scikit-image, IPython Notebook, ChatGPT
Languages
Python, C++, SQL
Paradigms
Data Science, Dynamic Programming, Database Design
Industry Expertise
Teaching
Platforms
Amazon Web Services (AWS), Azure, Docker, Google Cloud Platform (GCP)
Storage
Amazon S3 (AWS S3), PostgreSQL, Databases
Frameworks
Flask, CherryPy
Other
Machine Learning, Deep Learning, Computer Vision, Image Processing, Random Forests, Mathematics, Algorithms, Data Structures, Object Detection, Artificial Intelligence (AI), Data Analysis, Decision Trees, Probability Theory, Statistics, Convolutional Neural Networks (CNN), Data Analytics, Computer Vision Algorithms, Research, Labeling, Neural Networks, Startups, Intelligent Character Recognition, Defect Detection, Natural Language Processing (NLP), Kalman Filtering, OCR, Tesseract, Data Engineering, Linear Regression, Object Tracking, 3D Image Processing, Image Search, Number Theory, PySQL, Loss Modeling, Open Neural Network Exchange (ONNX), Visualization, Video Processing, Information Extraction, Graph Theory, FAISS, Pytesseract, 3D Reconstruction, SVMs, Deep Sorting, Image Recognition, Statistical Methods, Statistical Modeling, OpenAI, Training
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring