Viktor Tóth
Verified Expert in Engineering
Machine Learning Engineer and Developer
Viktor polished his machine learning and software skills in a wide range of projects, from petabyte-level data analysis and inference in the cybersecurity field to deep learning computer vision segmentation on light-microscopy images of peripheral nerves in the field of neuroscience. He's studied and gained professional experience in science and engineering in the academy and industry. Viktor excels at developing deep learning algorithms in Python and writing high-performance C++ software.
Portfolio
Experience
Availability
Preferred Environment
Linux, Python, C++, PyCharm, Git
The most amazing...
...project I've handled was the design of a deep learning model to translate images into sound for the blind, relying on psychoacoustics and brain imaging studies.
Work Experience
Machine Learning Researcher
Looka Inc
- Researched latent diffusion models and adjacent deep learning models and techniques to generate images from text prompts with reliable text rendering on images.
- Developed data pipelines, custom-designed diffusion model architectures, and fine-tuned latent diffusion models to generate images with reliable text rendering without relying on massive computational resources.
- Implemented then combined deep learning techniques from machine learning publications and custom-edited the stable diffusion architecture to manipulate its attention networks and to feed in extra visual information relevant to text rendering.
Machine Learning Researcher
Peafowl Plasmonics AB
- Researched optimization methods (e.g., genetic algorithms and bayesian optimization) to find parameters to a chemical synthesis process. We are publishing our solution.
- Implemented a Bayesian optimization algorithm that arrived at the optimal parameters in a matter of hours instead of weeks compared to the previous solution.
- Developed a graphical user interface used by chemical engineers that orchestrated the optimization process and the related hardware, i.e., chemical pumps and a spectrometer.
Machine Learning Engineer
SignalPET, LLC
- Researched, tested, and integrated deep learning computer vision networks for diverse classification, semantic, and instance segmentation problems in TensorFlow (Keras) and Python.
- Optimized hyperparameters and pre-trained backbones on classification and detection problems. Improved model accuracy of locating pathologies on X-ray images by greater than 20%. Further boosted image processing and data pipeline performance in TF.
- Integrated semantic and instance segmentation and image quality assessment pipelines into the codebase. Deep learning models and image processing was run on AWS instances.
Neuroengineer
Feinstein Institutes for Medical Research
- Implemented a data pipeline and algorithms to map a peripheral nerve by performing instance segmentation on light microscopy images using deep convolutional networks and computer vision techniques, achieving over 95% detection accuracy.
- Built, published, and deployed a clinical prediction model (RNN)—enabling 25% of hospital patients to sleep without interruptions, trained on measurements of 2.3 million patients. Won second place in the innovation challenge at Northwell Health.
- Implemented multiple COVID-19 patient prediction models using causal and survival modeling, XGBoost, and linear models to foresee mortality in the vicinity of ventilation and to predict outcomes in other clinically relevant situations.
- Built a rodent virtual reality setup from scratch—which included the mechanics, hardware, and software—and taught rats to play a computer game in it; also designed and conducted the corresponding animal experiments.
- Implemented a data pipeline for an implanted stereo EEG analysis—designing statistical measures to establish between-subject brain response consistency analysis to vagus nerve stimulation.
- Managed multiple summer students in successful, published deep learning, computer vision, and statistical analysis projects.
Software Engineer
Sentido Emotions Tech
- Developed real-time computer vision algorithms for emotion recognition in C++.
- Optimized an emotion detection algorithm to run on multiple threads in parallel enabling real-time processing of video streams.
- Researched psychological measures relevant to the recruitment use case of emotion detection.
Data Science Intern
Synack
- Designed and implemented a data pipeline and algorithms to cluster and compare hundreds of thousands of web pages to detect changes over time that may introduce cybersecurity vulnerabilities.
- Optimized the DBSCAN clustering algorithm for a specific use case on MapReduce (in Hadoop and Google Cloud Platform), achieving a three-fold running time reduction.
- Improved a tree matching algorithm that compared HTML trees to accurately detect changes in webpages over time, gaining a six-fold running time improvement.
- Implemented an anomaly detection data pipeline and algorithm to expose the atypical activity on close-to-a-million endpoints using machine learning techniques (hierarchical temporal memory from Numenta).
- Contributed to a patent: Automatic Webpage Change Detection, US20180191764A1.
Software Engineer
eRAD
- Improved performance and upgraded codebase to C++11 of eRAD's "photoshop for radiologists" software.
- Developed a mobile interface to a web service in the Unity engine.
- Implemented a Windows software installment, a deployment process for their C++ program.
Software Engineer Intern
National Centre for Biological Sciences
- Optimized the MOOSE biological neural simulation engine by implementing parts of the voltage propagation computation on the GPU (CUDA).
- Profiled Python C++ extensions (Cython) using Gperftools to measure bottlenecks of the neural simulation environment.
- Implemented a graphical user interface for the parameter optimization tool of the neural simulation engine.
Project Leader | Game Developer
Kitchen Budapest
- Led a team of four (two software engineers, a game designer, and an artist) to develop a mobile game where AI algorithms compete.
- Developed a C# mobile game using the Unity game engine.
- Implemented AI algorithms that were hand-tweaked by the user through a high-level, intuitive interface, before they battled each other in real time.
Experience
Analysis of Implanted EEG Recordings
I further performed non-negative matrix factorization on the evoked spectrograms to reveal shared features of brain activity between electrodes and subjects.
Peripheral Nerve Mapping
https://ieeexplore.ieee.org/document/9175974I combined a custom computer vision algorithm with a deep convolutional instance segmentation network to achieve an over 95% accuracy in fiber detection.
Tools Used: Gaussian mixture models, Voronoi tessellation, Detectron2 Mask-RCNN, geometric manipulations
Deep Learning Clinical Prediction Model
I developed a machine learning model to predict whether a patient's health is likely to deteriorate overnight using a deep recurrent neural network (LSTM) trained on sequences of 26 million vital sign measurements, e.g., heart rate, temperature, blood pressure, and so on.
Patients who are highly unlikely to deteriorate then don't receive unnecessary overnight vital monitoring—enabling more than 25% of patient nights to be left undisturbed.
The model was published in Nature Digital Medicine and is being deployed in multiple hospitals.
Translating Vision into Sound for the Blind
https://medium.com/mindsoft/translating-vision-into-sound-443b7e01ecedThe network training is engineered to produce sound modulations that are audible for humans, so a small visual difference in the image will result in a slight but discernable sound shift. Psychoacoustical and brain imaging studies supported all the assumptions embedded in the sound modulation. I blindfolded myself for five days without an interruption to test the algorithm and successfully learned to translate visual information into sound in two separate tasks.
Cybersecurity Web Change Detection
https://patents.google.com/patent/US20180191764A1/enI created a custom DBSCAN algorithm in MapReduce that clusters web pages according to their locality-sensitive hash (SimHash) by building a search tree out of the simhashes. I then compared the DOM of web pages of sufficiently similar simhashes using a custom optimized tree edit distance algorithm. Next, I reduced the runtime of DBSCAN by 4x and scaled up to hundreds of thousands of web pages on computing clusters which sped up tree edit distance computation by 6x.
After the web pages are clustered and compared, their differences (e.g., an added input text field) are surfaced to employed penetration testers to assess whether a vulnerability is introduced.
Thoughtflow: Dl Search Engine Matching People of Similar Instant Thoughts
https://medium.com/mindsoft/thoughtflow-io-connecting-people-by-sentence-sentiment-clustering-45f0b5f1203I hosted the deep learning model on GCP and provided a web interface to search and chat handling requests in Flask.
Skills
Languages
C++, Python, Java, C#, R, SQL, Assembly, C
Libraries/APIs
TensorFlow, PyTorch, Pandas, XGBoost
Tools
PyCharm, Git, Slack, MATLAB, Visual Studio, Cloud Dataflow
Paradigms
Data Science, MapReduce, Parallel Programming, Anomaly Detection
Platforms
Linux, NVIDIA CUDA, Apache Pig, Google Cloud Platform (GCP), Arduino, Amazon Web Services (AWS)
Other
Software Architecture, Machine Learning, Signal Processing, Deep Learning, Computer Vision, EEG, Artificial Intelligence (AI), Cluster Analysis, Recurrent Neural Networks (RNNs), Convolutional Neural Networks (CNN), Statistical Analysis, Data Visualization, Research, Image Processing, Image Analysis, Neural Networks, Algorithms, Mathematics, Data Analysis, Image Generation, Diffusion Models, Bayesian Statistics, Reinforcement Learning, Complexity Theory, Emotion Recognition, Startups, Neuroscience, Technical Writing, Natural Language Processing (NLP), Data Mining, Generative Pre-trained Transformers (GPT), Text Generation, Analysis of Variance (ANOVA), Windows Deployment, Profiling, Neuroinformatics, Simulation Engines, Machine Learning Deployment, Causal Inference, Modeling, Experimental Design, Computational Neuroscience, Optimization, Audio Engineering, Deployment, Web Development, Clustering Algorithms, Genetic Algorithms, Artificial Neural Networks (ANN), Bayesian Inference & Modeling, GUI Design, Chemistry, Simulations, Object Detection, GPT
Frameworks
Unity3D, Hadoop, Flask
Industry Expertise
Cybersecurity
Storage
MySQL
Education
Master's Degree in Human Neuroscience and Technology
Aalto University - Helsinki, Finland
Bachelor's Degree in Computer Engineering
Budapest University of Technology and Economics - Budapest, Hungary
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring