Kavan Alipanahi Email & Phone Number

Tehran Province, Iran

• Element Counting: Implemented TREX and Grounding DINO models to count objects on a production line conveyor belt from video data, achieving 4% higher accuracy than manual operation. • Voice Classification: Utilized mel spectrograms to convert audio signals into images and applied CNNs to classify the sounds with high accuracy, achieving over 95% recall in detecting broken components and enabling real-time component status monitoring.• Fire and Smoke Detection: Developed a multi-threaded fire detection system using YOLOv8, connecting each thread to a camera to track fires in real-time and trigger alarms with location information, achieving 100% detection accuracy. • Plastic Injection: Developed an OpenCV-based rule method on Raspberry Pi to detect stuck components in plastic injection molding, achieving over 98% accuracy and enabling autonomous operation without any human intervention.

Vancouver, British Columbia, Canada

• Led a team of 5 in AI-driven medical image segmentation and preprocessing with PyTorch.• Utilized PyTorch for segmentation of CT and PET images with cross-modal registration using diffusion models and Evidential UNet.

Tehran, Iran

• Golestan Co: Collaborated with the latest versions of YOLO (YOLOv8) to detect real-time data of store shelves and identify 8 different brands of nuts. Achieved over 90% mAP (Mean Average Precision) on real-time data, leading to an impressive 80% reduction in questionnaire costs at Golestan Company.• Nasim Group: Scraped 40,000 SIM card numbers along with their prices from the rond.ir website and built machine learning models to determine the prices and implemented a system to receive notifications of the most valuable prices based on the model's error.• Shahid Beheshti University: Instructed and trained undergraduate computer engineering students in Machine Learning for two months.

Tehran, Iran

• Analyzed data and built sales strategies in connection with over 800 IT companies, resulting in a 30% increase in the company's satisfaction.• Designed a machine learning model to analyse employers' behaviour in applying to jobs based on job post features, and created a "how to get more CVs" guide for our customer companies. • Designed contests for data science enthusiasts that feature real-world challenges with over 1300 participant.• Lead the sales team in engaging with customers.

Tehran, Iran

• Petro Sadaf: Developed a mobile application for participating petrol stations for facilitating their daily work schedule.• Maniban: Developed Mani Ban Application You can easily make your purchases by using your credit from our numerous recipients including restaurants, Coffeeshops, and Sports clubs.The Intelligent Search Engine of the App will show you the nearest recipients Quickly.

Paradise Hub

•Developed an application for managing family loan funds and building charges, which solves the challenges of accounting and tracking members for the fund manager and its members.