Serhan Narlı
PhD candidate at Charité Berlin · Machine learning engineer · Animal gait and lameness analysis
I work on computer-vision–based gait analysis for animals, applying deep learning to biomedical and clinical contexts. Building end-to-end ML pipelines that work in farms, clinics, and real-world environments.
About
I'm a PhD candidate at Charité – Universitätsmedizin Berlin working on deep learning for biomechanical analysis and early lameness detection in animals. My research focuses on developing computer-vision systems that can automatically assess gait patterns and identify subtle signs of lameness in dairy cattle.
My background spans computer science, signal processing, and medical AI. I'm particularly interested in the intersection of deep learning and biomechanics, where we can leverage modern AI techniques to solve real clinical problems.
I focus on building end-to-end ML pipelines that actually work in farms, clinics, and real-world environments – not just in controlled lab settings. This means developing robust systems that can handle challenging lighting conditions, camera angles, and the messy reality of production environments.
Research & Publications
I work on computer-vision–based gait analysis and medical AI, with a focus on automated lameness detection and reproducible pipelines.
Automated detection of lameness in dairy cattle through computer vision analysis of back shape characteristics
Computers in Biology and Medicine · 2025
Back-shape–based lameness detection pipeline for dairy cattle. We extract curvature features along the spine and train ML models to distinguish healthy vs. lame cows, comparing performance to human locomotion scorers.
Validation of a Deep Learning Model for Cattle Lameness Detection: Comparison of Human Scorer Performance and Automated Gait Analysis
Preprint · 2025
Validation study comparing a deep-learning gait model against multiple human raters, analysing inter-rater agreement and conditions under which automated scoring can match or exceed human consistency.
DeepCOVIDNet-CXR: Deep learning strategies for identifying COVID-19 on enhanced chest X-rays
Biomedical Engineering / Biomedizinische Technik · 2024
CNN-based approach for COVID-19 recognition on enhanced chest radiographs, exploring preprocessing and architecture variants to improve diagnostic performance.
Selected Projects
A mix of research prototypes and applied AI systems.
Automated Cattle Gait Analysis Platform
Research & development
Web-based analysis platform for dairy farms. Users upload walking videos of cows and receive automatic locomotion scores. The system detects keypoints along the back, computes curvature and gait features, and predicts lameness risk.
Jetson-Based Real-Time Gait Monitoring System
Edge AI pipeline
Farm-ready edge pipeline running on NVIDIA Jetson for real-time cow monitoring. GStreamer-based camera + recording service with YOLO analysis, RFID triggers and custom data-quality metrics for keypoints.
Historical Figures Interactive Chat App
Generative AI product
Conversational AI platform that lets users talk to simulated historical figures while enforcing factual consistency and strict personas through prompt design and retrieval-augmented generation.
DeepCOVIDNet-CXR
Medical imaging model
Deep learning pipeline for detecting COVID-19 on enhanced chest X-rays using custom preprocessing and CNN architectures.
EEG-Controlled 6-Axis Robot Arm
B.Sc. thesis
Prototype system that maps EEG signals to control commands for a 6-axis robotic arm, exploring low-cost brain–computer interfaces.
Writing & Essays
I write about neuroscience, artificial intelligence and how technology shapes our future.
The History of EEG
A long-form essay on how EEG evolved from early electrical experiments to modern quantitative EEG and brain–computer interfaces.
Read on Medium →A Philosophical Approach to Machine Learning
An introduction to machine learning and neural networks from a philosophical angle – how the brain learns, how perceptrons work, and why making mistakes is essential.
Read on Medium →Will Neuralink Really Hack Our Brains?
A critical look at Neuralink: what brain–machine interfaces can actually do today, why 'mind hacking' is still science fiction, and where the real progress lies.
Read on Medium →EEG Sinyal İşleme ve Derin Öğrenme
Turkish article on recording and processing EEG signals with deep learning, based on an Erasmus project on emotion recognition using Emotiv headsets.
Read on Medium →Experience
AI/ML Engineer
MCG Motion Capture GmbH, Berlin
Developing production-ready computer-vision features, Jetson-based pilots and integrating AI tools into commercial motion capture products.
Signal Processing Engineer Intern
StepUp Air Solutions, Copenhagen
Built signal-processing pipelines for wearable respiratory sensors and real-time monitoring.
Research Intern
Wroclaw Technical University
Worked on computer-vision and autonomous systems prototypes as part of academic collaborations.
Education
PhD in Machine Learning (ongoing)
2022 – presentCharité – Universitätsmedizin Berlin
Focus on deep learning for biomechanical analysis and early lameness detection in animals, combining video-based gait analysis with clinical validation.
M.Sc. in Computer Science Engineering
2017 – 2019İskenderun Technical University
Thesis on deep learning for medical image analysis and biomedical signal processing.
B.Sc. in Computer Science Engineering
2012 – 2017İskenderun Technical University
Developed an EEG-controlled 6-axis robotic arm and multiple embedded systems projects.