Azmi is available for hire
Hire AzmiAzmi Can Ozgen
Verified Expert in Engineering
AI/ML Engineer and Developer
Istanbul, Turkey
Toptal member since May 14, 2026
Bio
Azmi is an AI/ML engineer with 8+ years of experience building data science solutions, AI and ML pipelines, and computer vision systems. He has a strong research background with many academic publications. Azmi is comfortable with reporting and presenting projects at different levels.
Portfolio
Venhancer
Data Science, Machine Learning, Computer Vision, Large Language Models (LLMs)...
Confidential client
Data Science, Machine Learning, Python, PyTorch, NumPy, Pytest, Pandas, Polars...
Epiforest AB
Data Science, Machine Learning, Computer Vision, Python, NumPy, OpenCV, Pandas...
Experience
- Python - 10 years
- Linux - 10 years
- Programming - 10 years
- OpenCV - 9 years
- Computer Science - 9 years
- Computer Vision - 9 years
- Scikit-learn - 8 years
- PyTorch - 6 years
Work Experience
Senior AI Engineer
2026 - PRESENT
Venhancer
- Supervised and directed about 6 mid-level and junior developers and interns across client projects; joined client meetings and ran technical hiring (interviewed and onboarded an intern).
- Researched and benchmarked local LLMs (Qwen, Gemma) for Viennalife to replace GPT models in a live system that transcribes client calls with Whisper and fills life-insurance forms with an LLM; supervising a junior developer on the benchmark.
- Researched a contextual bandit layer over an LLM chatbot that selects the most informative next question to recommend car insurance products for Sigortam.net (TÜBİTAK R&D project).
- Implemented and presented a computer vision demo for car crash detection, for Sahibinden.
- Coordinated a financial document comparison product that matches client finance records against Integra's records using NLP and LLMs, for Integra.
Technologies: Data Science, Machine Learning, Computer Vision, Large Language Models (LLMs), Python, Whisper, PyTorch, Qwen, Google Gemma, OpenAI GPT-4 API, Claude Code, Linux, Bash, Git, FastAPI, Next.js
Senior AI Engineer
2026 - 2026
Confidential client
- Built a gradient-boosting (LightGBM) correction pipeline with clustering-based post-processing that substantially reduced estimation error versus the baseline approach.
- Engineered signal and statistical features to improve model accuracy.
- Established a rigorous cross-validation framework for honest out-of-sample generalization and diagnosed model failure modes.
- Built an internal web tool for signal annotation to support dataset labeling.
Technologies: Data Science, Machine Learning, Python, PyTorch, NumPy, Pytest, Pandas, Polars, Matplotlib, Seaborn, Flask, Jupyter, Scikit-learn, Linux, Bash, Git, LightGBM, CatBoost, Claude Code
Senior AI Engineer
2026 - 2026
Epiforest AB
- Developed a terrain analysis pipeline that classifies harvesting terrain into branch extraction, ground protection, and impassable zones from slope and soil-moisture raster, with direction-dependent cross-slope hazard detection to prevent soil damage.
- Built an automatic zone-detection algorithm that partitions a forest tile map into 2-6 terrain zones by minimizing within-zone difficulty variance, then generates and scores multiple main/side track layout options per zone.
- Implemented a coverage route optimizer using vehicle-routing (Clarke-Wright savings) with A* path search, producing multiple strip-pattern options (vertical, horizontal, diagonal, flow-following) for expert selection.
- Developed a Gradio annotation tool for manual graph editing and route definition, and a synthetic dataset generator producing verified 512x512 georeferenced samples (EPSG:3006) for training and validation.
- Presented weekly progress and incorporated project-manager feedback into the terrain classification and track-scoring model.
Technologies: Data Science, Machine Learning, Computer Vision, Python, NumPy, OpenCV, Pandas, Polars, Matplotlib, Seaborn, Jupyter, Scikit-learn, Linux, Bash, Git, Streamlit, Gradio, Claude Code
Senior ML Engineer
2024 - 2026
Grid Dynamics
- Developed a defect classification model in an imbalanced class setting.
- Created and deployed an ML pipeline suitable for Databricks and Azure environments.
- Presented the weekly data science work to the stakeholders.
- Assisted the deployment team with bug fixing and improving the ML pipeline.
- Adapted the ML pipeline for new production lines and new data sources.
Technologies: Python, PyTorch, Scikit-learn, OpenCV, Weights & Biases, MLflow, Databricks, Azure, Machine Learning, Data Cleaning, Data Cleansing, Statistical Analysis, Image Processing, Artificial Intelligence (AI), NumPy, AI Model Training, ML Pipelines, Fine-tuning, GPU Computing, K-means Clustering, Ubuntu, Image Analysis, Data Labeling, Object Detection, Image Classification, Web Scraping, Data Visualization, Feature Engineering, You Only Look Once (YOLO), Machine Learning Operations (MLOps), NVIDIA CUDA, SciPy
Data Scientist
2025 - 2025
The Plum Guide
- Developed a home quality scoring system with at least 85% precision for the 3-type quality metric B, A, and A*.
- Developed a data labeling Streamlit interface for image annotation, which improved the annotation speed at least 5x.
- Deployed a microservice system for the entire pipeline using Docker.
- Presented new pipeline results and performances to the project managers and reported the progress weekly.
Technologies: Python, PyTorch, OpenCV, Streamlit, Docker, Prometheus, Azure, Data Cleaning, Data Cleansing, Machine Learning, Dashboards, Artificial Intelligence (AI), Image Processing, Pytest, Pandas, Polars, Microservices, Linear Regression, NumPy, SQL, AI Model Training, ML Pipelines, Fine-tuning, GPU Computing, K-means Clustering, Ubuntu, Image Analysis, Data Labeling, Object Detection, Image Classification, Regression, Beautiful Soup, Data Visualization, Feature Engineering, NVIDIA CUDA, SciPy
Senior Computer Vision Engineer
2021 - 2024
REM People
- Developed SKU detection and segmentation models for retail products, which performed with at least 90% F1-score.
- Developed a price tag recognition system with 90% precision and 85% recall for price tag detection and 95% precision and 90% recall for price text recognition.
- Guided another team in developing an image stitching system for retail shelf images, resulting in an enhancement of shelf analysis performance by a minimum of 10%.
- Published a paper on a price tag recognition system at the Signal Processing and Communications Applications Conference (SIU) 2022.
Technologies: Python, PyTorch, OpenCV, Weights & Biases, Gradio, Docker, Google Cloud Platform (GCP), Large Language Models (LLMs), Machine Learning, Natural Language Processing (NLP), Artificial Intelligence (AI), Optical Character Recognition (OCR), Vector Databases, Data Cleaning, Data Analysis, Image Processing, Image Segmentation, Pandas, Facial Recognition, NumPy, AI Model Training, ML Pipelines, Fine-tuning, TensorFlow, GPU Computing, K-means Clustering, Amazon Web Services (AWS), Ubuntu, Image Analysis, Data Labeling, Object Detection, Image Classification, Data Visualization, Feature Engineering, You Only Look Once (YOLO), Machine Learning Operations (MLOps), NVIDIA CUDA, SciPy
AI Engineer
2020 - 2021
Huawei Technologies Co.
- Developed a cheating detection system for online interviews with at least 90% precision and 85% recall for cheating activity detection.
- Developed a flagging system for suspicious activities in online interviews to speed up the review process for human reviewers.
- Published a paper on the cheating detection system at the Signal Processing and Communications Applications Conference (SIU) 2021.
- Deployed the Python-written pipeline as an executable service for a Java-based application using PyInstaller, which improved the deployment speed and reduced the maintenance cost.
Technologies: Python, PyTorch, OpenCV, FFmpeg, Linux, PyInstaller, Facial Recognition, Video Processing, Image Processing, Machine Learning, Artificial Intelligence (AI), NumPy, AI Model Training, Fine-tuning, TensorFlow, GPU Computing, Time Series Analysis, Ubuntu, C++, Image Analysis, Object Detection, Image Classification, Video Analysis, Feature Engineering, NVIDIA CUDA
ML Engineer
2018 - 2019
Miletos
- Contributed to the development of a bank check recognition system by training, evaluating, and optimizing several text recognition models.
- Contributed to the development of a car damage analysis system for insurance claims by reporting the results of models.
- Contributed to the development of an Ottoman handwriting recognition system by researching and applying new algorithms and reporting the results of models.
Technologies: Python, PyTorch, OpenCV, Google Cloud Platform (GCP), Docker, Linux, Image Processing, Data Analysis, Data Cleaning, Optical Character Recognition (OCR), Pandas, Machine Learning, NumPy, AI Model Training, Fine-tuning, TensorFlow, GPU Computing, K-means Clustering, Time Series Analysis, Support Vector Machines (SVM), Ubuntu, C++, Image Analysis, Data Labeling, Object Detection, Image Classification, Data Visualization, Feature Engineering, NVIDIA CUDA
Project History
Online Interview Cheating Detection System Implementation
https://github.com/azmiozgen/text-detectionA cheating-detection system for online interview processes of candidate videos. It has a flagging system for suspicious activities, such as seeking help from someone else or using a mobile phone. It was deployed as a Python application as an executable service for a Java-based system.
Words as Art Materials: Generating Paintings with Sequential GANs
https://github.com/azmiozgen/text2paintingThe generation of artistic images on a large variance dataset. This dataset includes images with variations, for example, in shape, color, and content. These variations in images provide originality, which is an important factor for artistic essence. One major characteristic of our work is that we used keywords as image descriptions, instead of sentences. As the network architecture, we proposed a sequential generative adversarial network (GAN) model. The 1st stage of this sequential model processes the word vectors and creates a base image, whereas the next stages focus on creating high-resolution artistic-style images without working on word vectors.
Geometry Friends AI
https://github.com/azmiozgen/geometry-friends-aiGeometry Friends is a 2D platform game developed at GAIPS INESC-ID, http://gaips.inesc-id.pt/geometryfriends/.
This project presents a game environment for developing game-playing AI. An AI agent is trained by using reinforcement learning and inspired by the works of Yanpanlau for the game of Flappy Bird.
This project presents a game environment for developing game-playing AI. An AI agent is trained by using reinforcement learning and inspired by the works of Yanpanlau for the game of Flappy Bird.
Education
2017 - 2020
Master's Degree in Computer Engineering
Istanbul Technical University - Istanbul, Türkiye
2011 - 2016
Bachelor's Degree in Physics
Istanbul Technical University - Istanbul, Türkiye
2011 - 2016
Bachelor's Degree in Electronics Engineering
Istanbul Technical University - Istanbul, Türkiye
Certifications
AUGUST 2026 - PRESENT
Claude Code 101
Anthropic Education
AUGUST 2026 - PRESENT
Claude 101
Anthropic Education
JULY 2023 - JULY 2025
Google Cloud Certified Professional Data Engineer
Google Cloud
FEBRUARY 2020 - PRESENT
Introduction to Data Science in Python
Coursera
FEBRUARY 2018 - PRESENT
Neural Networks and Deep Learning
Coursera
JUNE 2015 - PRESENT
Introduction to Computational Thinking and Data Science
edX
MAY 2015 - PRESENT
Introduction to Linux
edX
MARCH 2015 - PRESENT
Introduction to Computer Science and Programming Using Python
edX
Skills
Libraries/APIs
PyTorch, OpenCV, NumPy, FFmpeg, Gradio, Scikit-learn, Pandas, TensorFlow, SciPy, Beautiful Soup, Natural Language Toolkit (NLTK), Matplotlib, CatBoost
Tools
You Only Look Once (YOLO), PyInstaller, Pytest, Claude, Claude Code, Seaborn, Jupyter, Git, Whisper
Languages
Python, C++, SQL, Bash
Platforms
Linux, Ubuntu, NVIDIA CUDA, Weights & Biases, Docker, Google Cloud Platform (GCP), Azure, Databricks, Amazon Web Services (AWS)
Frameworks
Streamlit, Flask, LightGBM, Next.js
Paradigms
Microservices
Other
Engineering, Programming, Computer Science, Computer Vision, Data Science, Machine Learning, Image Processing, Artificial Intelligence (AI), Image Segmentation, Linear Regression, AI Model Training, Fine-tuning, GPU Computing, K-means Clustering, Data Labeling, Object Detection, Image Classification, Data Visualization, Feature Engineering, Prometheus, MLflow, Data Analysis, Polars, Optical Character Recognition (OCR), Facial Recognition, Generative Adversarial Networks (GANs), Support Vector Machines (SVM), Dimensionality Reduction, Decision Trees, Logistic Regression, Random Forests, Image Analysis, Regression, Video Analysis, Electronics, Physics, Data Engineering, Neural Networks, Deep Learning, Convolutional Neural Networks (CNNs), Data Processing, Data Cleaning, Data Cleansing, Statistical Analysis, Dashboards, Large Language Models (LLMs), Natural Language Processing (NLP), Vector Databases, Reinforcement Learning, Deep Reinforcement Learning, Game AI, Game Development, Game Design, Video Processing, ML Pipelines, Time Series Analysis, Web Scraping, Machine Learning Operations (MLOps), Qwen, Google Gemma, OpenAI GPT-4 API, FastAPI
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring