Felix B
Data Scientist / Machine Learning Specialist
Experience
Machine Learning Developer
Machine status detection in industrial 3D printing based on infrared image data
Baseline: Client has infrared cameras installed inside production machines for manual monitoring. Goal: Automate the process monitoring to identify irregularities from live images. Activities: I contributed to this project as a machine learning developer and project coordinator:
- Guided systematic data collection and pre-processing for the machine learning algorithms
- Defined the labeling process and implemented an interface to annotate the datasets
- Programmed a visual deep learning algorithm to detect machine pollution in live production
- Implemented data augmentation techniques to deal with machine heterogeneity
- Supply a containerized model with API endpoints for deployment to the production machines
- Coordinated and represented a five-person project team, prepared presentations and reports
Outcome: Provided a production-ready system to inform operators of potential faults in real time.
Technologies: PyTorch, Convolutional neural networks, docker, git, OpenCV, FastAPI, computer vision, pytorch-lightning, ipywidgets
Machine Learning Developer
Quality prediction based on time series data in a manufacturing context
Baseline: Multitude of sensors collects time series data from production machines at high frequency. Goal: Identify patterns in the data linked to final product quality. Activities: I worked in this project as a machine learning developer and project coordinator:
- Analyzed and visualized large amounts of time series data
- Explored, filtered and connected various data sources to construct consistent datasets
- Conceptualized the data pipelines and modeling approaches
- Implemented a machine learning system for quality prediction in a manufacturing context
- Tested various predictive algorithms including statistical models and deep neural networks
Outcome: Derived qualitative insights to guide process engineers in optimizing production.
Technologies: PyTorch, CNNs / ROCKET model, data wrangling, git, time series classification, explainable AI
Machine Learning Scientist
dida Machine Learning
- Conducted deep learning research projects in computer vision and time series
- Implemented commercial machine learning software in a manufacturing context
- Developed data processing pipelines for the deployment into live production
- Conceptualized application of large language models for question answering
- Represented and organized a five-person project team
- Prepared analyses and visualizations of large amounts of unstructured data
Machine Learning Scientist
Automated question answering via retrieval of internal documents
Baseline: Company employees manually search internal documentation for information. Goal: Integrate chatbot for question answering (QA) into company communication platform. Activities: I supported this project as a machine learning scientist:
- Created a demo for extractive QA based on a provided document
- Developed proof-of-concept for automatic question answering (QA) with semantic search
- Expanded processing pipeline to larger quantities of internal documents via automatic document retrieval and generative QA with large language models (LLMs)
- Informed the developers about recent advancements in natural language processing (NLP)
Outcome: Made responsive chatbot for document-based extractive QA available to all employees.
Technologies: transformers, haystack, LLMs, document retrieval, semantic search, question answering, BERT, git, beautifulsoup4
Econometrics Researcher
Estimation and analysis of variance spillover networks for academic research
Baseline: Existing network estimation methodology only works reliably on small datasets. Goal: Apply machine learning methods to analyze large datasets of economic and financial data. Activities: I realized this project as an econometrics researcher:
- Explored and compared statistical learning algorithms for multivariate forecasting
- Acquired datasets from SQL databases and set up an automated pre-processing pipeline
- Wrote object-oriented code to run cross-validated regularized machine learning algorithms
- Conducted extensive statistical and econometric analyses to empirically analyze the results
- Authored research papers that present the results at academic standard
Outcome: Published acclaimed paper in Quantitative Economics, follow-up papers under review.
Technologies: pandas / numpy, time series forecasting, scikit-learn, python, SQL, glmnet, econometrics, vector auto-regression, networkx
Teaching Assistant
Universidade Nova de Lisboa
- Supported teaching and grading of Master's courses in financial data analysis
- Generated course materials, assignments and exams
Quantitative Strategies Intern
risklab / Allianz Global Investors
- Provided data-driven research on fund performance and portfolio optimization
- Prototyped tools for investment strategies in Matlab and VBA for Excel
- Modeled time series with regularized regressions and scenario simulations
Project Management Intern
zeb Consulting
- Prepared senior management presentations to guide an IT project in risk management