Florijan Stamenković
Verified Expert in Engineering
Machine Learning Developer
Zagreb, Croatia
Toptal member since August 1, 2018
Since 2004, Florijan has worked on a wide range of projects such as machine-learning platforms for large Croatian and Austrian sites, a high-performance database engine in C++, client-server systems in Java and more. Besides having strong technical skills, Florijan excels at translating a client's ideas and needs into fully functional software and has experience in project management and leading teams.
Portfolio
Experience
- Machine Learning - 15 years
- Python - 15 years
- Data Science - 10 years
- Artificial Intelligence (AI) - 10 years
- Natural Language Processing (NLP) - 10 years
- Generative Pre-trained Transformers (GPT) - 10 years
- Deep Learning - 8 years
- Computer Vision - 8 years
Availability
Preferred Environment
C++, Python, Data Science, Machine Learning
The most amazing...
...project I've led and implemented was a real-time, personalized recommendation system for the largest Croatian media company.
Work Experience
Computer Vision Expert
Undisclosed
- Conceived and implemented a computer vision pipeline for the processing of macro photography previously done by human experts.
- Tuned multiple processing algorithms based on the client's proprietary dataset and optimized algorithm performance.
- Deployed computer vision processing APIs on AWS, utilizing Lambda, S3, and EC2.
- Supported the development of a front-end application for using the developed pipelines.
MLOps Engineer via Toptal
Brain Craft Limited
- Designed and implemented an ML job execution platform based on a set of custom-developed services, AWS, and cloud GPU providers.
- Utilized Docker and Docker Compose for the development and production deployment of multiservice stacks to support ML model usage.
- Integrated multiple GPU cloud provider platforms (notably Vast.ai and RunPod) into a self-monitored, automated system.
- Prepared several models for production deployment (text-to-video generation, object detection, and image inpainting).
- Optimized the performance of image processing services through image preprocessing and model tuning.
- Coordinated two junior team members who were developing support for other ML models based on the same stack.
Data Scientist via Toptal
Roompulse A.E.
- Developed and optimized price estimation and anomaly detection algorithms in the hospitality domain.
- Used OpenAI GPT models for text analysis, processing, and inference.
- Worked and advised on data gathering and cloud pipeline approaches.
Cloud Architecture Engineer
Pangea Botanica (via Scalexa)
- Designed, developed, and deployed a job execution platform on AWS supporting massive parallel execution, GPU execution, on-demand scaling, map-reduce, queuing, etc.
- Developed and integrated data acquisition, processing, and versioning pipelines for multiple public chemical databases.
- Implemented APIs and CLI tools for platform interactions (job execution and data acquisition).
Computer Vision Developer
Inci Islim
- Used OpenCV to detect augmented-reality markers in the video.
- Implemented algorithms for improved marker tracking resilience.
- Implemented visualizations and annotations over existing videos.
Machine Learning Engineer
Kesef Global Pty Ltd
- Optimized Large Language Models (LLMs) on a specialized dataset using state-of-the-art models and tech (Hugging Face models, Accelerate, DeepSpeed).
- Worked on model reduction (quantization, pruning) and training speed optimizations.
- Analyzed the Bittensor platform codebase (nodes, datasets) and suggested development directions based on findings.
Data Scientist | Machine Learning Engineer
Procureship S.A.
- Developed a recommendation engine based on previous procurement data.
- Wrapped the recommendation engine into an easily deployable web service.
- Educated the client about the recommendation-engine fundamentals, data requirements, and basic algorithms and advised on future development.
Machine Learning Engineer
Microblink
- Researched current state-of-the-art recommendation systems for eCommerce and worked on several implementations of such systems.
- Analyzed purchase data, testing various premises supporting buyer behavior predictions.
- Used TensorFlow and Docker to develop encapsulated machine learning services.
Computer Vision Researcher | Engineer
Photomath
- Proposed and developed multiple prototype systems in computer vision.
- Researched existing state-of-the-art computer vision papers and implemented some of them.
- Developed and deployed neural network-based computer vision tools and classic computer vision (OpenCV) algorithms.
Machine Learning Engineer
Velebit AI
- Evaluated papers on portrait segmentation and chose them among competitive approaches regarding available data and requirements.
- Worked on network architecture and model training. Implemented data processing pipelines (preprocessing, augmentation, and visualization). Prepared tooling for data labeling.
- Evaluated papers on object detection and adapted algorithms to a specific domain.
Python Developer | Data Engineer
Labelbox
- Designed and implemented the Labelbox Python SDK, open-sourced at Github.com/Labelbox/labelbox-python, and available on PyPI (pip install labelbox).
- Worked on ETL pipelines using the Google Cloud Platform (BigQuery and Data Studio) for analytics.
- Worked on massive parallel processing using Apache Beam (GCP Dataflow) for image processing.
- Used Apache Airflow (GCP Cloud Composer) for job scheduling (ETL and periodic recalculations).
- Worked on IAM integrations and automation for GCP and AWS.
- Implemented queue-based (RabbitMQ and Celery) data processing pipelines.
NLP Engineer
QiO
- Implemented a semantic search engine tailored for a specific domain (database of documents).
- Implemented an event extraction (from natural language), supporting a platform for converting unstructured to structured data.
- Applied topic modeling and text summarization algorithms.
Data Science Consultant
Picks and Pickers S.L.
- Evaluated the current state of the social network in Picks and Pickers and recommended an approach for improving engagement.
- Evaluated client and product data regarding the possibilities of making a product recommendation engine.
- Analyzed the product database for possibilities of deduplication and product description similarity matching and merging.
- Implemented a text classifier using Scikit-learn and Flask.
Machine Learning Consultant
Sensie
- Supported a wellness application for stress detection based on mobile device sensor data.
- Implemented and evaluated machine learning classifiers based on motion sensor data.
- Deployed machine learning models using Google Cloud Function.
- Analyzed existing classification algorithms and advised on the future development.
Information Retrieval Engineer | Analyst
Piggy | Joinpiggy.com
- Implemented custom information retrieval (search) algorithms based on NLP and ML using Sklearn and Numpy.
- Implemented a custom database search algorithm based on textual query ranking and geographical closeness, using Sklearn and Scipy.
- Optimized service for low latency and high throughput workloads using Numpy and Cython. Prepared the deployment using Flask.
- Provided analytics of user behavior data in numeric and visual form using DynamoDB, Pandas, and Matplotlib.
- Implemented data-reduction, aggregation, and anonymization pipelines for delivery to third parties.
Data Scientist for Cost Regression
ARTA Shipping, Inc.
- Implemented a regression model for a fine-art shipping cost estimation using Pandas and Sklearn.
- Performed data analysis using Pandas. Generated reports and visualization of the data using Matplotlib.
- Implemented a prediction service utilizing the optimized model using Flask.
- Implemented a training, reporting, and deployment pipeline for continuous integration.
Software Engineer on a Transactional Database Engine
Memgraph
- Worked on a transactional, distributed, in-memory graph database engine focused on high-performance using modern C++ (14 and 17 standards).
- Worked on a query compiler, including parsing, syntactic and semantic analysis, intermediary codes, interpretation, and compilation.
- Built a distributed database engine version with dynamic sharding and balancing capabilities.
- Worked on a multi-version concurrency control (MVCC) based transaction engine for a parallel query execution with snapshot guarantees.
- Worked on concurrent data structures and algorithms.
- Developed in a very-high standard environment: continuous integration, in-depth code reviews, unit/integration/compliance testing, and performance regression analysis.
Technical Lead
Styria Data Science
- Acted as the technical data science lead during the growth of a newly founded machine learning department from three to more than ten employees.
- Prototyped, developed, and deployed machine-learning software for some of the largest Croatian (24sata.hr and Njuskalo.hr) and Austrian (Willhaben.at) sites.
- Built NLP models for real-time, personalized recommendations with a microservice stack around it. Used neural networks, autoencoder stacks, word and paragraph embeddings, and more.
- Developed computer vision technology for image classification, similarity, and object detection, using traditional computer vision methods and neural networks (deep learning).
- Handled GPU deployments, both locally and on AWS. Worked on S3 data pipelines, data processing automation, model, and runtime speed optimizations.
- Acted as the technical evaluator in the hiring process for developers and interns.
Software Engineer
CNG Havaso Ltd.
- Worked as the lead programmer in a small startup, focusing on production-site data management.
- Developed a modular client-server system for issue-tracking, inventory, and bookkeeping.
- Implemented the back end using Java, Apple's WebObjects, and Enterprise Objects frameworks, backed by Openbase and Frontbase.
- Implemented a desktop application connecting to the Web Objects app server using Java Swing.
- Developed and open-sourced JBND, a Swing data-binding library for rapid application development.
- Gave a talk at the worldwide WebObjects conference in San Francisco (2008) on JBND and desktop client apps.
Experience
Memgraph
https://memgraph.com/Labelbox Python SDK
https://github.com/Labelbox/labelbox-pythonWillhaben Fashion Camera
I worked as the tech lead of the development team, overseeing the model implementation and evaluation, data pipelining (large volumes of images transferred between Austrian hosting, AWS Ireland, Germany, and Croatian hosting), search-engine runtime implementation, and product development.
Object Detection in 3D
https://www.youtube.com/watch?v=v5ubMhAki_QPersonalized Recommendater for 24sata.hr
https://www.24sata.hr/Regression Modeling for ShipArta
https://shiparta.com/Sensie Stress Detection
Education
Master's Degree in Computer Science
University in Zagreb - Zagreb, Croatia
Bachelor's Degree in Computer Science
University in Zagreb - Zagreb, Croatia
Certifications
Sun Certified Java Programmer
Oracle
Skills
Libraries/APIs
Pandas, TensorFlow, Scikit-learn, PIL, PyTorch, NumPy, OpenCV, SQLAlchemy, Theano, REST APIs, Keras, AMQP, Google API, Google APIs, NetworkX, Pika, SciPy, Matplotlib, Standard Template Library (STL), SpaCy, Protobuf, OpenGL, DeepSpeed
Tools
Pytest, GitHub, Git, TensorBoard, Phabricator, GIS, Jupyter, ChatGPT, pylint, Apache Beam, Apache Airflow, Cloud Dataflow, Google Analytics, BigQuery, Vim Text Editor, Boto, AWS IAM, GeoNames, ANTLR, RabbitMQ, Celery, Scikit-image, Amazon Elastic Container Service (ECS), Docker Hub, Docker Compose, uWSGI, AutoCAD, AWS Fargate
Languages
Python 3, SQL, Python, C++, Cypher, C++17, GraphQL, Java SE, HTML, Bash, Java, C, Objective-C, C++14, Bash Script
Paradigms
Object-oriented Programming (OOP), Test-driven Development (TDD), Unit Testing, ETL, DevOps, Agile, Microservices, REST
Platforms
Docker, Google Cloud Platform (GCP), Amazon Web Services (AWS), Amazon EC2, Linux, Web, AWS Lambda, Kubernetes
Storage
Relational Databases, Data Pipelines, PostgreSQL, Google Cloud, Amazon S3 (AWS S3), Databases, Neo4j, MySQL, Redis, MongoDB
Frameworks
Flask, Google Test, WebObjects, Enterprise Objects Framework (EOF), gRPC
Industry Expertise
Project Management
Other
Artificial Intelligence (AI), Clustering, Convolutional Neural Networks (CNNs), Image Processing, Data Analytics, Recommendation Systems, Data Analysis, Data Science, Computer Vision, Natural Language Processing (NLP), Deep Learning, Machine Learning, Data Visualization, Object Detection, Machine Learning Operations (MLOps), Data, Analytics, Large Language Models (LLMs), Computer Vision Algorithms, Data Engineering, Neural Networks, Predictive Modeling, Back-end, Generative Pre-trained Transformers (GPT), Graphics Processing Unit (GPU), Multithreading, APIs, Search, Algorithms, Cython, Concurrency, Multiprocessing, GraphDB, Computer Graphics, Product Leadership, Big Data, Technology Consulting, Object Tracking, Search Engines, Data Modeling, Data Extraction, API Documentation, Cloud, CI/CD Pipelines, Language Models, Fine-tuning, Infrastructure, Application Performance Monitoring, Google Data Studio, Google Cloud Functions, Google BigQuery, Heuristics, Genetic Algorithms, Image Search, Objects, Storage, Optical Character Recognition (OCR), Compilers, ARM, Geospatial Data, Point Clouds, User Behavior, Serverless, API Integration, Open Source, Statistics, Causal Inference, Data Inference, Text Generation, Scripting, Video Processing, Revenue Optimization, Pricing Strategy, Chatbots, OpenAI GPT-4 API, Abstract Syntax Trees (AST)
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