Milan Panic

I provided technical leadership for the Evercheck project for more than five years. Evercheck is designed to reduce shrinkage in retail stores by targeting the leading causes of retail loss, particularly at self-checkout stations and cashier-assisted registers.

The system addresses over 30 unique fraud and loss patterns, with key use cases including detection of non-scans, where it identifies both intentional and accidental missed scans. It also detects product switching, triggering alerts when a product's barcode has been swapped or covered with that of a less expensive item. Additionally, Evercheck recovers losses at self-checkout by identifying unscanned items left inside or at the bottom of shopping carts or baskets.

Evercheck generates real-time alerts for store associates, enabling them to assist customers promptly when unscanned items are detected, thereby enhancing loss prevention and improving the shopping experience.

I provided consultancy services in deep learning, system design, and project organization, contributing to the development of Allimb—a virtual healthcare professional designed to assist patients during clinical recovery and physical exercise.

Allimb offers real-time feedback and correction of complex body movements without requiring any external equipment or internet connection.

I provided technical and strategic leadership across neural network and framework development, Android and iOS application development, and the creation of AI-powered assessment tools.

One of the most meaningful projects I contributed to was PainChek, an innovative solution designed to give a voice to individuals who cannot verbalize their pain. I applied AI technology to support a modern pain assessment framework capable of detecting pain that may not be immediately visible.

Using a smartphone's camera, PainChek captures images of a person's face and analyzes them with AI-driven facial recognition algorithms. It automatically identifies and records facial muscle movements that indicate pain, making the tool fast, accurate, and easy to use in real-world clinical settings.

I led the system design and development of the initial framework for neural network training focused on facial expression analysis. The early use case centered on recognizing individuals' expressed emotions, with the goal of enabling automatic analysis of viewer reactions to advertisements.

This foundational work later contributed to the development of BeEmotion, now operating within Discovery AI. Discovery AI is at the forefront of intelligent solutions, leveraging advanced algorithms and machine learning to build adaptive applications. It supports a wide range of domains—including Smart Mobility, Smart Living, and Smart Health—and serves as a platform for studying, testing, and validating new approaches in human-machine interaction and intelligent sensing.

I led the development and ongoing management of the full digital presence (excluding social media) for Modern Talking, an online foreign language school.

This long-term project includes:
• Main website: moderntalking.rs (WordPress, JavaScript)
• Cross-platform app: platform.moderntalking.rs for web, iOS, and Android (React Native, MySQL, MongoDB, Flask, Python)
• AI-powered content: Custom tools built with Hugging Face, Google AI Studio, and TTS technology to generate learning materials from course content.

AI free samples:
• AI-generated Narration and Lesson Summary: https://www.notion.so/mtalkingschool/Lesson-1-All-that-you-can-do-with-only-one-ZDRAVO-216d2658eebe80a69128fa9ed7b050d8

• AI Podcast Episode - https://www.youtube.com/watch?v=-921G9Z-0pM