Pavol Mulinka

Collaboration on open source deep learning library pytorch-widedeep:

• Deep Imbalanced Regression
• New loss functions
• Custom Imbalanced DataLoader

Collaboration on Data Science related tasks under the username Pavol86, please see the Phabricator tickets and related git merges:

• Compress Gensim models
• python-mwtext
• Tokenization of ”word” things for CJK
• deltas
• revscoring
• editquality

Analysis and classification of data from diverse security tools and systems. Pattern, anomaly, and potential security threats detection. Research and application of machine learning methods (NLP, LLM, statistics) to enhance and enrich solution outputs.

Design and analysis of machine learning approaches for pattern and anomaly detection in synthetic & real-world data. For related projects see the Projects section. Related projects:

• SUCCESS-6G: Towards robust, secure and computationally efficient vehicular services in 6G
  • git repo - https://5uperpalo.github.io/success6g-edge/
  • description - data analysis, design, and implementation of the solution for AIoT in V2X
  • used technologies - Python(numpy, pandas, matplotlib, etc.), MySQL, MLflow, Kubeflow, Kubernetes(Minio, Prometheus, Zero-to-Jupyterhub, Flower, Kserve, Knative, Istio, Kepler, Grafana, InfluxDB), streamlit, models(LLMs/Huggingface, Transfrormers, GBMs, sklearn, torch, pytorch-widedeep)
• FIREMAN (Framework for the Identification of Rare Events via MAchine learning and IoT Networks)
  • git repo - https://github.com/5uperpalo/FIREMAN-project
  • description - data analysis and implementation of the solution for classification of rare events in IIoT
  • used technologies - Python(numpy, pandas, matplotlib, etc.), models(LLMs/Huggingface, Transfrormers, GBMs, sklearn, torch, pytorch-widedeep, HDBSCAN, OPTICS, DBSCAN, GANs, GAIN, scikit-multiflow/river, deep-river), MOA, Kafka, Airflow, MLflow, KSQL, Faust, InfluxDB, Flask

Design, analysis and impl. of machine learning approaches for (i) prediction of customer lifetime value in mobile apps and (ii) in-app purchase recommendation system. Related projects:

• IAP (In-App-Purchases)
  • description - data analysis, design and implementation of recommendation system for In-App-Purchases on mobile phones
  • used technologies - Python(numpy, pandas, matplotlib, etc.), Athena/SQL, AWS, models(LLMs/Huggingface, Transfrormers, GBMs, sklearn), Weights&Biases
• PLTV (Predicted LifeTime Value)
  • description - data analysis, design and implementation of solution to predict life time value of customers from aggregated data from their early in-app behavior, and additional auxiliary data
  • used technologies - Python(numpy, pandas, matplotlib, etc.), Athena/SQL, AWS, models(LLMs/Huggingface, Transfrormers, GBMs, sklearn, torch, pytorch-widedeep, H20, DBSCAN, OPTICS), Weights&Biases,

Design and development of communication interface between Slovak Electricity Hydro optimization model and user GUI. Docker containerized solution utilizing Redis for in-memory database, Flask as web framework and Celery for multi-processing. Design of code unit testing. Transformation of procedural code to object oriented. Code refactorization.

Migration of existing Apache XML firewall/loadbalancer solution to F5 loadbalancer for a Ministry of agriculture project

• Network Engineer (VSHosting, Prague, CZ)
• Network Consulting Engineer (Verizon, Prague, CZ)
• Senior System Engineer (AT&ampT, Bratislava, SK)
• HP Radia Specialist (Soitron, Bratislava, SK)
• HP Monitoring Support Specialist (Soitron, Bratislava, SK)
• IT VoIP support specialist (Soitron, Bratislava, SK)

Application of the unsupervised machine learning (ML) approaches to network (NW) traces (MAWI, Darknet). Generalization and improvement of the hierarchical density-based clustering approach to NW measurements interpretation proposed during AIT Vienna internship. Improvement of PySpark ML scripts running in distributed UX server environment. Results were summarized in conference papers.

Analysis of the relations between socioeconomic status of customers and network performance, and investigation of potential discrimination in network deployment and management. Correlating LSOA database (Lower-layer Super Output Areas) and operator measurements by Geographic Information System (QGIS, ArcGIS, GeoPandas) in distributed computing env. (PySpark).

Cybersecurity and network performance analysis, anomaly detection and diagnosis. Application of supervised, unsupervised, batch and stream-based machine learning techniques on big network measurement datasets (MAWI and Cloud latency). Integration of machine learning approaches into big data analytics platforms - in particular, working on a distributed computing environment within the BIG-DAMA project. Utilization of distributed computing tools and platforms such as Cloudera, PySpark, Apache Pig, Hive, Kafka, Elasticsearch etc.. Running and configuration of machine learning bash script on linux server. Results were summarized in conference papers.

Mentoring, academical support and provisioning of computing environment for under-grad intern visiting NII Tokyo for three weeks supported by Sakura Science Plan internship.

Network operating systems, Linux, Unix. Administration and network tools, managing and administration of documentation. Basic concepts, configuration and procedures in operating systems administration (UNIX).

Basic principles of both classical and programmable logic devices and their practical use in the design of digital systems. Design and implementation of digital circuits VHDL language. Implementation of logic gates, measurement of their static and dynamic properties. Verification of digital circuits in the simulator.

Review of switching systems solution principles, i.e. (i) switching fields, (ii) control systems and (iii) signalization for switching control (in central office as well in networks). Focus on digital switching systems with circuit commutation as well as transport of IP packets. Basic review and consideration about convergence of voice and data services and networks including functional principles of new generation networks with respect to philosophy and services of intelligent network.