Mateusz Dalba

Objectives: To reduce the response time for common customer queries and to automate repetitive support tasks.

Data: Jira, Instructions, Confluence, any relevant documentation in various formats.

Modeling: Fine-tuned GPT models using OpenAI. Prompt Engineering. Custom ML models trained with HuggingFace and Scikit-Learn.

Technology: NLP, OCR, Python, OpenAI, Flask, Streamlit, HuggingFace, Numpy, Scikit-Learn, Azure Cloud.

Challenges: Poor data quality. Handling diverse customer queries, ensuring seamless integration with existing systems, and maintaining high accuracy in responses.

Outcomes: Significant reduction in response time, increased customer satisfaction, and freed up support team members for more complex tasks.

Objectives: Streamline workflows, reduce manual effort, and increase overall efficiency.

Data: CSV, Excel files, MySQL, Snowflake, data analytics methodologies extracted from Jupyter Notebooks.

Modeling: Supervised ML, Unsupervised ML, Neural Networks, Statistics, Advanced Data Analytics.

Technology: Python, Scikit-Learn, Streamlit, Django, Keras, Matplotlib, Seaborn, Plotly, Pandas, Scipy, AWS Cloud.

Challenges: Dynamically changing methodologies. Ensuring compatibility with existing workflows and handling diverse data sources.

Outcomes: Increased productivity, reduced errors, and more time for data scientists to focus on complex analysis.

Objectives: To enhance the accuracy of energy demand forecasts and optimize resource allocation.

Data: Records of historical demand.

Modeling: Time Series Analysis and Forecasting. Regression models. Neural Networks. Statistics.

Technology: Python, Pandas, Streamlit, Flask, Azure Cloud, Sklearn, Statsmodels, Skforecast, Plotly, Seaborn, Matplotlib, Scipy.

Challenges: Forecasts evaluation, XAI (models interpretability). Dealing with seasonality and trends, integrating with existing systems.

Outcomes: Developed application was embedded into main forecasting process in demand planning team. Improved demand forecasting accuracy, better resource management, and cost savings.

Objectives: To enable the shopping cart to recognize and track products autonomously.

Data: Images of products taken from shopping cart camera.

Modeling: Object detection, object recognition, semantic segmentation, Deep Neural Networks, YOLO.

Technology: Python, OpenCV, Keras, Tensorflow, Flask, PIL, YOLO, Numpy, Skimage, Matplotlib, Pandas.

Challenges: Lack of additional camera to get the "depth" of cart. "Wild" environment and dynamic nature of data (supermarket). Real-time image processing, handling various lighting conditions, and ensuring high accuracy in product recognition.

Outcomes: Enhanced shopping experience, reduced checkout time, and increased customer satisfaction.

Objectives: To accurately detect Polish banknotes in real-time.

Data: Self gathered dataset of Polish banknotes. Created artificial samples of data.

Modeling: Multiple custom trained Deep Neural Networks.

Technology: Python, OpenCV, Numpy, Tensorflow, Keras, Pandas, Matplotlib, Skimage, Tflite.

Challenges: Hard edge cases, managing different conditions of banknotes and handling various lighting conditions.

Outcomes: Blind people can recognize what types of banknotes they have in their wallets.

Objectives: To analyze race data and provide actionable insights for strategy optimization.

Data: Telemetry data. Car sensors data. Data taken from simulators and real cars. Expert knowledge.

Modeling: Supervised ML. Unsupervised ML. Neural Networks. Statistics.

Technology: Streamlit, Flask, Python, Scikit-Learn, Keras, Tensorflow, Numpy, Matplotlib, Seaborn, Plotly, Scipy.

Challenges: Transfering knowledge from simulators to real life scenarios. Real-time data analysis, handling large volumes of data, and integrating with racing telemetry systems.

Outcomes: Enhanced racing strategies, improved race performance, and data-driven decision making.

Objectives: To enhance online shopping by providing an immersive fitting room experience, enabling users to visualize how clothing items will look on them before making a purchase.

Data: VITON dataset. Human faces datasets (e.g. CELEBa). Human postures datasets. Data scrapping.

Modeling: Semantic segmentation. Generative Adversarial Networks (GANs). Human key-points detection neural networks.

Technology: Python, Tensorflow, Keras, OpenCV, Skimage, Numpy, Cloud computing.

Challenges: Lack of data. Accurate try-on models prediction with least amount of artifacts.

Outcomes: Increased user engagement, reduced return rates, and improved online shopping experience.

Objectives: To improve support operations by extracting and analyzing keywords and semantic similarities between projects, as well as to gauge customer sentiments and perceptions through feedback analysis.

Data: Project documents, support tickets, customer feedback, surveys, and social media comments.

Modeling: NLP modeling - semantic similarity, sentiment analysis, keywords extraction.

Technology: DataBricks, Python, PySpark, HuggingFace, Numpy, Scikit-Learn, Keras, Tensorflow.

Challenges: Managing and analyzing diverse and large volumes of data, ensuring accurate semantic analysis, and integrating insights into existing systems for actionable results.

Outcomes: Enhanced support efficiency through faster case resolution and relevant case matching, improved understanding of customer sentiments, actionable insights for business strategies, and better customer relationship management.

Objectives: To provide a comprehensive introduction to data analytics and data science principles and practices.

Technology: Python, Jupyter Notebooks, Data Visualization, Pandas, Numpy, Scikit-learn, Tensorflow, Keras, Streamlit.

Challenges: Creating engaging content, ensuring accessibility, and covering a wide range of topics comprehensively.

Outcomes: Positive feedback from learners, and increased understanding of data science concepts.

Objectives: To tackle diverse data science challenges and deliver practical solutions.

Technology: Python, R, Data Visualization tools, and Machine Learning libraries, Data Bricks, Azure Cloud, AWS Cloud, Google Cloud.

Challenges: Handling diverse project requirements, managing limited resources, and ensuring timely delivery.

Outcomes: Successful project completions, satisfied clients, and practical solutions to data-related problems.