Tomislav is available for hire
Hire TomislavTomislav Ivek
Verified Expert in Engineering
Data Science Developer
Location
Zagreb, Croatia
Toptal Member Since
September 2, 2022
Tomislav is a bilingual professional with a PhD in condensed matter physics. He has a strong background with over 15 years in research and 5 years in the industry as a consultant dedicated to developing cutting-edge AI technologies. An accomplished project manager and communicator, Tomislav makes his mark in designing tailor-made solutions based on the latest breakthroughs in machine learning.
Portfolio
Institut za fiziku Zagreb (Institute of Physics)
Physics, Artificial Intelligence (AI), Python, PyTorch, Deep Neural Networks...
Freelance Clients
PyTorch, Python, R, Data Analysis, Data Modeling, Deep Neural Networks...
Institut za fiziku Zagreb (Institute of Physics)
Python, Optics, Mentorship, Publication, Time Series, Data Analysis, Research...
Experience
Availability
Part-time
Preferred Environment
Linux, Python, Visual Studio Code (VS Code), Docker, PyTorch, Jupyter Notebook, TensorBoard, NumPy, Pandas, Artificial Intelligence (AI)
The most amazing...
...things I've done include managing world-class research projects and applying computer vision to fight climate change.
Work Experience
Senior Research Associate
2021 - PRESENT
Institut za fiziku Zagreb (Institute of Physics)
- Managed a 9-person international team as the lead of a competitive research project.
- Reviewed research project proposals within the European Union.
- Collaborated on an award-winning neural model to predict the likelihood and severity of wildfires. The inpainting-based computer vision model handles more than 40% of missing data and scores better than state-of-the-art statistical frameworks.
- Participated in organizing international scientific conferences and workshops.
- Mentored two doctoral students in experimental physics, instrument control, data collection, and advanced statistical methods.
- Published over 40 original research papers on Web of Science, covering hot topics in physics, biophysics, and extreme value statistics.
Technologies: Physics, Artificial Intelligence (AI), Python, PyTorch, Deep Neural Networks, Team Management, Mentorship, Lecturing, Publication, Public Speaking, Time Series, Docker, Jupyter Notebook, Data Analysis, Research, Experimental Research, Linux, Statistics, Writing & Editing, Convolutional Neural Networks (CNN), Statistical Modeling, Expert Reviews, Parallel Programming, GPU Computing, Data Cleaning, Object Detection, Object Recognition, Scikit-learn, LaTeX, Data Science, Deep Learning, Generative Adversarial Networks (GANs), Machine Learning, Climate Change, Neural Networks, Git, Data Visualization, Python Asyncio, REST, Open Source, Predictive Modeling, Reviews, Artificial Neural Networks (ANN), Data Modeling, Team Leadership, Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), GPT, Reports, Forecasting, AI Design, Signal Processing, Digital Signal Processing, Python 3, Fine-tuning, Data Inference, Text Generation, DeepSpeed, Algorithms
Consultant and Deep Learning Developer
2017 - PRESENT
Freelance Clients
- Evaluated viable computer vision approaches and implemented a solution for reliable object detection, classification, and tracking in a noisy, domain-specific video input, reaching a mean average precision (mAP) of above 0.97.
- Researched the topic of deep learning in financial forecasting, then implemented, trained, and launched the resulting model for live data testing.
- Compared a conventional recommender system with a custom deep learning model, then designed and trained an ensemble recommender system.
- Implemented a robust system for the distributed training of a pharma deep AI model, beating state-of-the-art models in its accuracy in detecting trace amounts of target matter in the measured signal.
- Debugged an astrophotography stacking app and contributed code to develop nonconventional optical sensors.
Technologies: PyTorch, Python, R, Data Analysis, Data Modeling, Deep Neural Networks, Artificial Intelligence (AI), Computer Vision, Generative Adversarial Networks (GANs), Computer Vision Algorithms, Artificial Neural Networks (ANN), Deep Learning, Time Series, Time Series Analysis, Rust, Financial APIs, Financial Data, Forecasting, Trend Forecasting, AI Design, Data Scraping, Web Scraping, Signal Processing, Digital Signal Processing, Quantitative Finance, Fintech, Python 3, Causal Inference, Fine-tuning, Data Inference, Language Models, Text Generation, DeepSpeed, Datasets, Scraping, Algorithms, Recommendation Systems
Research Associate
2015 - 2021
Institut za fiziku Zagreb (Institute of Physics)
- Managed two successful bilateral EU projects and one national research project in experimental natural sciences with a diverse international team of nine researchers and students.
- Devised a neural method to predict oceanographic extreme temperature events and published the related paper working in a team. The model leverages a denoising autoencoder similar to recent Stable Diffusion and inpaints missing ocean temperatures.
- Wrote a compatibility shim for Python GPIB and VISA libraries to control measurement equipment and ease porting between Windows and Linux. It allowed three laboratories to transition their data acquisition code toward Linux and open-source software.
Technologies: Python, Optics, Mentorship, Publication, Time Series, Data Analysis, Research, Experimental Research, C++, Linux, Statistics, Writing & Editing, Linear Regression, LabVIEW, Deep Neural Networks, GPT, Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), GPU Computing, Data Cleaning, Artificial Intelligence (AI), Scikit-learn, LaTeX, Physics, PyTorch, Data Science, Generative Adversarial Networks (GANs), Machine Learning, Climate Change, Statistical Analysis, Git, SQL, Classification, R, Raspberry Pi, Computer Vision Algorithms, Data Visualization, DSP, Open Source, Predictive Modeling, Spreadsheets, Student Engagement, Reviews, Data Modeling, Team Leadership, Reports, Data Scraping, Web Scraping, Signal Processing, Digital Signal Processing, Python 3, Physics Simulations, Scraping, Algorithms, Qt
Postdoctoral Researcher
2012 - 2014
University of Stuttgart
- Analyzed time-dependent infrared spectra correlated with fast electric transport measurements.
- Implemented PID control of temperature and sample position on a multi-user cryogenic measurement set up.
- Mentored and oversaw doctoral and master students between two collaborating groups.
Technologies: Fourier Analysis, Optics, Fiber Optics, Python, Lecturing, Mentorship, Publication, Instruments, Time Series, Data Analysis, Research, Experimental Research, Statistics, Writing & Editing, LabVIEW, Statistical Modeling, Data Cleaning, LaTeX, Physics, Team Management, Data Science, Statistical Analysis, Neural Networks, Data Visualization, DSP, Open Source, Spreadsheets, Student Engagement, Reviews, Data Modeling, Team Leadership, Reports, Signal Processing, Digital Signal Processing, Physics Simulations, Algorithms
Research Assistant
2005 - 2012
Institut za fiziku Zagreb (Institute of Physics)
- Graduated with PhD in experimental condensed matter physics on strongly correlated electron systems, including superconductors, charge orderings, strange metals, and semiconductors.
- Implemented an experiment control system for dielectric spectroscopy at low temperatures.
- Managed multiple websites for international research projects, conferences, and schools.
Technologies: Linux, LabVIEW, C, C++, Physics, Time Series, Publication, Instruments, Data Analysis, Research, Experimental Research, Statistics, Writing & Editing, Linear Regression, Statistical Modeling, Data Cleaning, LaTeX, Data Science, Statistical Analysis, SQL, Data Visualization, Open Source, Microsoft Excel, Spreadsheets, Data Modeling, Reports, Trend Forecasting, Signal Processing, Digital Signal Processing, Algorithms
Experience
Domain-specific Finetuning of a YOLOv5 Object Detector
A YOLOv5 object detector that I applied to a video input. I evaluated viable computer vision approaches and implemented a solution for reliable object detection, classification, and tracking in a noisy, domain-specific video input. The base model was pre-trained on COCO and finetuned to domain-specific datasets—including power grid components, art, and botanics—using multiple GPUs. The mAP reached above 0.97.
An Award-winning Wildfire Severity Prediction AI
https://github.com/Blackbox-EVA2021/CMIWAEAn AI that predicts the severity and frequency of wildfires across the USA, which won 1st place at the Extreme Value Analysis 2021 data challenge. As part of a two-person team, I engaged in all project phases. I researched, designed, planned, and implemented a probabilistic neural network—which handles incomplete training data using optimization techniques—and contributed to the final write-up.
Document Summarization Using GPT3
A GPT3-based summarization of user-supplied PDF files representing reports and research papers. I designed and implemented a simple Python API which takes user-supplied document files, extracts the text, and passes it for analysis to OpenAI's GPT3 model, which summarizes it into salient points.
Red Sea Temperature Extremes
https://github.com/BlackBox-EVA2019/BlackBoxA neural network that predicts the extremes of sea temperatures based on incomplete oceanographic data. As part of a two-person team, I prepared the data, designed, trained, and served the model, and visualized the result. The project won second prize at the Extreme Value Analysis 2019 data competition.
Predictive Modeling of Financial Data
A PyTorch-based multitask model that I trained and which exceeded specified metrics. During the project, I researched the topic of deep learning in financial forecasting, trained and served the resulting model, and integrated it with backtesting and live trading APIs. Details under NDA.
Next-gen Experimental Chemical Characterization
A project where I used machine learning to improve the specificity of chemical compound detection using commercial instrumentation. I spearheaded data augmentation procedures to increase the robustness of trained models and contributed to a multidisciplinary team currently working on launching the related startup. Details under NDA.
DNA and Protein and Proteome Classification
Researched natural language processing techniques to apply to DNA and proteome classifiers. I implemented a transformers-based neural model of DNA and proteome sequences in PyTorch, which outperformed traditional methods with a precision and recall of up to 97%.
Collective Dynamics in Multiferroic Materials
A large-scale research project, which I conceived and led, focused on novel compounds for sensors and information storage. I implemented a new measurement and data analysis technique for the electric polarization of materials in extreme conditions. I also coordinated between the local team and several partner institutions in France, Germany, Slovenia, Croatia, and Serbia. Additionally, I mentored doctoral students and postdocs and published research papers.
GPIB-ctypes Library
https://pypi.org/project/gpib-ctypes/A cross-platform Python bindings library for the NI GPIB and Linux GPIB C interfaces. I implemented and published a cross-platform library that, on Microsoft Windows and Linux, presents an API compatible with the Linux GPIB Python bindings. I also facilitated data acquisition and control of measurement equipment without relying on proprietary binaries and Python bindings, enabling three research laboratories to seamlessly transition to Linux-only solutions with little to no code modification.
Equipment Control Using Raspberry Pi
A Python-based Raspberry Pi (RPi) server that I designed and implemented for measurement equipment control and data collection. I utilized an RPi2 device to orchestrate data collection and serve the data to users (low-temperature experimental physics). I also reduced data collection time from about 5 hours to 45 minutes per experiment and significantly decreased the need for human intervention.
AUR Recipe for the Conan Package Manager
https://aur.archlinux.org/packages/conanPackaged and maintained Conan, the C++ package manager for Arch Linux, enabling the use of a modern build system on Arch-based distributions. I maintained several related Python dependencies on the Arch User Repository (AUR) and contributed bug reports and fixes upstream.
Astrophotography Stacking
Collaborated on an application to align, stack, and superpose photographs of astronomical objects. I applied OpenCV to find "sky landmarks" and contributed a non-Bayer filter for Fujifilm optical RGB sensors.
Scientific Papers
https://scholar.google.hr/citations?user=4PwGYvIAAAAJ&hl=enCoauthored multiple scientific papers in top-tier journals and conference proceedings related to experimental physics, machine learning, chemistry, and statistical modeling. The full list can be viewed on Google Scholar. I've also demonstrated extensive experience in public speaking, lecturing, presenting, and mentoring.
Scientific Reviews
Served as a project reviewer for funding agencies and a referee for top-tier scientific journals and conferences, including the French National Research Agency: ANR, Scientific Reports, Physical Review B – American Physical Society, Applied Physics Letters, The European Physical Journal, Plus, ECRYS, Molecular Crystals and Liquid Crystals, and Electrical Engineering.
Education
2005 - 2011
PhD in Experimental Physics
University of Zagreb Faculty of Science - Zagreb, Croatia
1999 - 2005
Master's Degree in Physics
University of Zagreb Faculty of Science - Zagreb, Croatia
Skills
Libraries/APIs
Pandas, PyTorch, Scikit-learn, REST APIs, OpenCV, Python Asyncio, OpenGL
Tools
LaTeX, Instruments, Git, You Only Look Once (YOLO), LabVIEW, Conan, Microsoft Excel, Spreadsheets
Languages
C++, Python, Python 3, C, Rust, R, SQL
Paradigms
Parallel Programming, Data Science, REST
Platforms
Linux, Jupyter Notebook, Arch Linux, Raspberry Pi, Docker
Storage
SQLite, PostgreSQL
Frameworks
Qt
Other
Data Analysis, Fourier Analysis, Research, Experimental Research, Statistics, Writing & Editing, Linear Regression, Convolutional Neural Networks (CNN), Natural Language Processing (NLP), Time Series Analysis, Expert Reviews, Deep Neural Networks, Computer Vision, Data Cleaning, Artificial Intelligence (AI), Physics, Team Management, Mentorship, Publication, Public Speaking, Fiber Optics, Time Series, Deep Learning, Machine Learning, Neural Networks, Statistical Analysis, Unsupervised Learning, Classification, Text Classification, Supervised Learning, Supervised Machine Learning, Computer Vision Algorithms, Data Visualization, Open Source, Predictive Modeling, Mathematics, Student Engagement, Artificial Neural Networks (ANN), Data Modeling, Scripting, Reports, Forecasting, Trend Forecasting, Data Scraping, Signal Processing, Digital Signal Processing, Fine-tuning, Data Inference, Datasets, Physics Simulations, GPT, Generative Pre-trained Transformers (GPT), Optics, Multiprocessing, Statistical Modeling, Object Detection, Object Recognition, Lecturing, Project Budget Management, Climate Change, DSP, Multithreading, Team Leadership, Bayesian Inference & Modeling, Financial Data, Generative Adversarial Networks (GANs), AI Design, Web Scraping, Causal Inference, Language Models, DeepSpeed, GPU Computing, Scraping, Algorithms, Recommendation Systems, Reviews, Financial APIs, Quantitative Finance, Fintech, Text Generation, Videos, Generative Pre-trained Transformer 3 (GPT-3), Documentation, Executive Summaries
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