Denis Milošević, PhD

Bridging ML research advances and production-scale AI systems

Research scientist and ML engineer specializing in large-scale AI systems, with expertise in computer vision and multimodal architectures, data-intensive applications, and distributed computing. Drawing from doctoral work in optimizing computer vision models for medical applications, now architecting end-to-end ML solutions - from data strategy to production systems - for complex real-world problems through systematic evaluation and empirical validation.

Research Focus

Drawing from research experience in medical computer vision, I develop robust ML architectures and systems that balance innovation with practical reliability.

Vision & Multimodal ML

• Vision architectures & optimization
• Cross-modal understanding
• Efficient model design & validation

Generative AI

• Generative model architectures
• Cross-modal generation
• Task-specific model adaptation

Research-Driven Engineering

• Systematic evaluation & metrics
• Data-backed design decisions
• Performance optimization

ML Systems Architecture

• End-to-end system design
• Data annotation pipelines
• Scalable deployment solutions

Selected Papers

Some of my publications in conferences and journals that I'd like to highlight.

View all papers →

Automated Estimation of Chronological Age from Panoramic Dental X-ray Images Using Deep Learning

2022

Systems with Applications (Impact Factor: 7.5)

This paper introduces a novel deep learning approach for automated age estimation from panoramic dental X-rays, achieving state-of-the-art accuracy for adult subjects with a median error of just 2.95 years, significantly outperforming established methods, most of which require manual measurements and require no dental damage or alterations.

A Comprehensive Exploration of Neural Networks for Forensic Analysis of Adult Single Tooth X-ray Images

2022

IEEE Access (Impact Factor: 3.4)

This study presents a comprehensive exploration of deep learning for analyzing individual tooth X-ray images to perform three key forensic tasks - age estimation, sex assessment, and tooth type determination. By creating one of the largest datasets in literature (86,495 annotated teeth), the research resulted in automated models that achieve performance equal or better performance than current manual forensic methods, while being fully automated, and without requiring perfect dental specimens.

Impact & Experience

Systems: Designed, built, and deployed ML systems processing tens of millions of queries weekly across computer vision and multimodal applications
Research: Published peer-reviewed research advancing the state-of-the-art in medical computer vision systems and efficient ML architectures
Scale: Spearheaded end-to-end ML initiatives from research proof-of-concepts to production systems serving non-trivial workloads

Contact

Want to talk about making interesting stuff? Feel free to reach out! :D

Connect

@_denmil on 𝕏