Machine learning / Data science / Academic research
Private research toolkit exploring EEG, signal features and classical machine learning through desktop, offline and local web components.
Original code-generated category illustration; not a product screenshot.
Private codebase; this page shares a general technical summary. Source files, real records and organization-specific information remain private.
A collection of experimental Python tools for EEG and signal-processing research. Desktop acquisition workflows, offline feature/model experiments and a local audio-upload prototype are separate components. The collection demonstrates research-code organization without publishing participant recordings, model files or claims of clinical accuracy.
When acquisition, feature extraction and classifier experiments span different tools, their requirements and limitations can become unclear. Code execution, prediction accuracy and clinical validity are separate questions. Local prototypes accepting uploads also need basic constraints on how filenames and output locations are handled.
Signal processing, feature extraction and development and evaluation of machine-learning prototypes for EEG-based yes/no intent classification.
Desktop acquisition, offline research scripts and the local audio-processing prototype are treated as separate responsibilities. The general workflow explores numerical signal features and classical classifiers. A review-branch change normalizes uploaded filenames and rejects unusable names. Synthetic tests focus on that file-handling behavior; they do not establish scientific validity or successful end-to-end processing of real recordings.
I worked on EEG-based yes/no intent classification in research aimed at alternative communication for people living with ALS. My contribution covered signal processing, feature extraction and development and evaluation of machine-learning prototypes. The toolkit represents the software side of these experiments; it does not provide a ready-to-use communication device or a clinically validated system.
The desktop component handles experimental signal streams, while offline scripts focus on numerical features and classifier experiments. Separating these modes makes it possible to examine a feature or model change independently of data acquisition. Live signal quality, hardware connections, participant protocols and model evaluation require separate validation; successful script execution is not evidence of end-to-end accuracy.
The local audio-upload prototype is separate from the EEG classification workflow. Normalizing filenames and rejecting unusable names limits writes outside the intended destination directory. Two synthetic regression tests address this file-handling behavior. Their count cannot be interpreted as EEG prediction quality, hardware compatibility or a clinical outcome.
Handle experimental signal streams in the desktop component
Explore feature extraction and models in offline scripts
Keep local audio-upload experiments in a separate web component
Evaluate each component against its own data and validation boundaries
Conceptual flow derived from project descriptions; not an application screenshot.
Private document-processing prototype exploring local PDF-to-table extraction through batch processing and a desktop interface, with explicit privacy boundaries.
Private logistic-regression research experiment exploring binary classification, ROC visualization and explicit input validation for tabular data.