Development and implementation of the graphic design of the home page of the ISiM Department (AGH) based on Next.js and headless Wordpress

In order to refresh the appearance of the cathedral website, together with two devs, I am preparing a headless WordPress instance with Next. The project assumed presenting the offer of the ISiM department, presenting the staff and a classic WordPress blog. The staging environment was prepared on the RetroLab infrastructure and resources were made available using docker and traefik (currently internal only).

Implementation and administration of virtualized lab infrastructure called Retrolab

For the needs of the faculty, I prepared and made available a testbed for applications built in the cloud native model, called RetroLab. The entire environment is based on virtualization using Proxmox. A three-node cluster was created, consisting of server-class machines, each equipped with a GPU. Users have at their disposal a Kubernetes cluster (3x worker VMs, each with a GPU) and virtual machines with Docker. The whole thing was secured using opnsense and the Authentik service. Access is via Wireguard VPN. Other implemented services include: balanced with keepalived DNS based on pihole, Traefik as a reverse proxy, TrueNAS (NFS and minio), outline wiki, Label Studio.

Multi-sensory platform for imaging and detecting threats occurring in areas with high dynamics of changes in environmental conditions, codename MUSE

The main goal of the project is to create a software and hardware platform using unmanned aerial vehicles to image and detect military threats on land and sea. In the project, I am responsible for designing software architecture, in particular with regard to the use of ML models, the purpose of which is to recognize and track military objects based on images from a UAV camera. Difficult operational conditions in terms of communication (range, interference) are also taken into account. The project uses software and hardware resources such as: Jetson AGX Orin, python, tensorflow, Angular, MLflow. Project no.: DOB-SZAFIR/01/B/038/04/2021

Development of an intelligent feed management system increasing the automation of e-commerce business

The project involved carrying out research and development work leading to the development of a product feed management system, allowing the Applicant to provide the same service, i.e. product feed management service. The feed management system developed in the project will be offered to customers in the SaaS cloud model. The product feed informs a given advertising system about what the most important product parameters should look like in the online store. In technical terms, it is a file in Excel, .txt, .csv or .xml format and contains detailed information about the products offered by the online store. In the project, I was responsible for developing and preparing the cloud infrastructure for system deployment. The GCP cloud was used, and in particular its two services (apart from the application hosting itself): BigQuery, used for sales data analytics, and VertexAI, used to deploy ML models. Together, these two solutions were used to build product feeds. I also participated in preparing an application for financing from the regional operational program of the Lesser Poland Voivodeship. Recruitment number: RPMP.01.02.01-IP.01-12-003/20.

Development of an innovative hybrid cyber-physical system enabling integrated management of steelmaking technology with the transport of liquid metal in the EAF–LF–CCM steel plant for a sequential melting system, in terms of optimizing the temperature of liquid steel in the CC tundish, in order to improve the quality of continuous ingots

The main goal of the Project is to carry out R&D work aimed at developing a globally innovative hybrid IT system integrated with the infrastructure of the EAF-LF-CCM steelworks for optimizing and modeling the production process of continuous ingots, i.e. a solution that significantly contributes to improving the quality of CMC products Poland sp. z o. o. In the project I was responsible for: 1. Ladle furnace detection model and recognition of their numbers. An ML model was built for object detection and numerical label recognition. Python and opencv and tensorflow libraries were used. The model was implemented in the form of a dotnet worker and a dll written in cpp. 2. Multi-criteria rule-based optimization of the production plan based on archived data (sql-server + linq) and data received online from the controller. The module was implemented as a dotnet web app (kestrel). Project no.: POIR.01.01.01-00-0996/19

Development of a continuous non-destructive testing system together with the development of software controlling the operation of the linear welding machine based on the analysis of the mechanical parameters of the linear welding

The aim of the project was to develop a modular cyber-physical system for monitoring and controlling the process of linear welding of rolled sheets. I dealt with four aspects of the project: 1. Interpolation of process parameters for further data analysis and optimization (influxdb + python). 2. Optimization of database queries for time series in Timescale db (postgresql + python). 3. Detection of the edges of welded sheet metal for optimal welding angle calculation (opencv + python). 4. Implementation of fuzzy logic models to predict optimal welding parameters during the process (cpp + python). The software was deployed in a containerized environment using Docker. Project no.: POIR.01.02.00-00-0091/19

Boguchwała Smart City – Dynamically reconfigured ecological and public safety monitoring system with detection of sources, emitters and separation of safety areas

The aim of the project was to improve ecological and public safety in the area of the city of Boguchwała. As part of the project, measurements and disciplinary actions were carried out in the field of air pollution, acoustic and electromagnetic disturbances and hydrological threats. The module for detecting noise sources and measuring the electromagnetic field intensity was constructed based on the RaspberryPi and Arduino (Libelium) platforms. The necessary services were implemented on the devices using the Kubernetes orchestrator. The system allows automatic implementation of software updates throughout the cluster (7 measurement points). Modules for monitoring the cluster condition using Prometheus were also prepared. A broader description of the work was published at the ICCS 2023 conference: http://dx.doi.org/10.1007/978-3-031-36030-5_41.

Development of innovative solutions for the automation of wagon loading and coke testing at Sorting Plant III using advanced databases and IT systems

The main goal of the interdisciplinary project was to develop the concept of a complete system enabling the automated process of loading and testing coke. The system has a modular structure consisting of components based on mechatronic solutions, visual identification, optimization and control software. One of the system's functionalities was the transfer of coke samples for further laboratory analyses, enabling unambiguous identification of the collected samples and the identification of the batch and wagon from which the sample was taken. This was possible thanks to the visual identification module used to recognize wagon side numbers and track their position during loading. Data read from the side walls of wagons will also be used in the wagon loading control system to optimize processes. The identification module was based on machine learning, where an application in Python was created using the tensorflow library for detecting and recognizing labels on wagons. The application was deployed using an additional tensorflow serving module. Project no.: POIR.01.02.00-00-0212/17-00

Development of an optimal logistics model and cooling model in the warehouse before the pickling line at the Cold Rolling Mill of ArcelorMittal Poland S.A.

Warehouse logistics in the area of storing hot-rolled coils before the line in the Cold Rolling Mill is usually a big challenge because it involves the need to solve an algorithmic problem, referred to in the literature as NP-hard. The main goal of the project was to develop a hybrid computer system that would provide a holistic approach to solving the above problems by designing and implementing an integrated material tracking, logistics and numerical simulation system. In this project I developed optimization algorithms for coil logistics using linear programming using Python. Project no.: POIR.01.02.00-00-0211/17-00

Software and hardware system for collecting and analyzing data to create rules for controlling a liquid pump

The aim of this project was to investigate the influence of patient-specific factors, such as identified chronic diseases, performance factors (e.g. DLCO, BMI), but also factors related to treatment: rehabilitation, course of surgery, on the course and duration of drainage. The following tasks were defined as part of the work: 1) Design, implementation and implementation of a database system for collecting selected data in the field of lung surgery and drainage - MS Access, 2) Data collection using designed applications - python, 3) Creating scripts transforming data for the technology selected for research, dedicated to statistical analysis - java, python, 4) Exploratory data analysis aimed at identifying prognostic factors in the model - R, python, 5) Creation of rules for controlling the liquid pump based on conclusions from statistical analyses. The work resulted in my master's thesis! :)

Nationwide computer network for education in Poland - case study

As part of the project, analytical work was undertaken regarding the implementation of the investment entitled: Internet for schools. This investment assumed providing each educational unit with an Internet connection with a fixed bandwidth. As a part of the work performed, the network architecture was analyzed, in particular checking whether it should be flat or hierarchical. For a specific, selected type of network architecture, its system parameters were determined and their impact on its functioning was analyzed. Using Google Maps API and Matlab, the technical architecture of the adopted solutions was presented, in particular from the point of view of network nodes, indicating their locations and implementation costs. Details can be found in the final report - monograph: https://www.researchgate.net/publication/328026093_Ogolnopolska_siec_komputerowa_edukacji_-_studium_przypadku.

Analysis and interpretation of information threats in a metal industry enterprise