Błażej – Python, MLOps, Tensorflow

Classification of signal from optic fiber (NDA restricts the rest of the information).

• Built an optic fiber signal classifier for a proof of concept project for NCBiR after successfully leading the plausible requirements gathering for it;
• Coordinated the process of data gathering for the next version of the classifier;
• Worked on building an on-premise machine learning platform for the AI needs of the company.
• As a result, the metrics of the classifier and its performance were improved significantly.

A platform for students, parents, and teachers to help others with homework questions. The website is intended to strengthen student's skills across subjects such as English, mathematics, science, and social studies.

Blazej's primary responsibility was to act as a tech lead for ML platform creation and deploy a dozen models to production.

• Architected and built a machine learning platform for NLP on AWS, following Google's MLOps Maturity Levels, including training & evaluation pipeline, model registry, monitoring, batch inference, online inference on Kubernetes, data labeling service;
• Developed an ETL for data anonymization, which served as a primary source of data for the platform mentioned above;
• Trained a transformer-based subject classifier model for an internal client;
• Successfully onboarded and mentored two junior members to the project.

The main aim of the project was to create models predicting the load of each stock in multiple locations across Europe and to build an analytical dashboard around it.

• Collaborated with a team of 6 on stock management forecasts PoC project, building an analytical dashboard in Dash;
• Implemented both the backend and frontend in Dash framework, making it fast and easy to look at both the current state of the stock and the predictions of our model;
• Presented the solution to the client, which rounded off the PoC phase successfully.

The project's main goal was to create a web app that automates test data creation for a mobile app project for one of the largest Australian banks. In the end, it was used for producing test data based on the team's specifications.

• Gathered project requirements from the client, and instead of manually creating packages of data for testing, Blazej proposed, architected, and implemented an automatic solution for the problem;
• Worked on both: the frontend and the backend side of the app;
• Provided solutions resulted in the fact that the QA team didn't rely on others to create test data and could automatically form whatever data they required via the application.