Selected projects
A few projects I have worked on.
They cover my master's research, API testing, computer vision, text analysis, and biosignals.
01
Causal ML | NLP | Research
The public companion to my master's thesis. I combined multilingual text analysis, channel-level holdouts, and causal ML to study crypto-market manipulation.
- Messages studied
- 12.4M
- Unseen-channel macro-F1
- 0.76
- Successful-event recall
- 0.89
Thesis results; the public repository uses synthetic fixtures and publishes no private Telegram data.
02
Agents | APIs | Evaluation
A Python tool for checking how an OpenAPI specification behaves when an agent uses the API through MCP. It validates the spec, runs test workflows, grades traces, and removes sensitive values from reports.
03
Computer vision | Remote sensing
A ResNet18 classifier for ten EuroSAT land-use classes, with a fixed data split and three preprocessing variants.
89.53% previously reported test accuracy
04
NLP | Streaming data
A streaming text-analysis tool for finding supplement mentions, questions, sentiment, negation, and aspects in Reddit discussions.
05
Biosignals | R/Shiny | ML
A course project with Paula Banach that combines an R/Shiny app with Python HRV preprocessing and WESAD-based stress experiments.
Educational research software. It is not a medical or diagnostic tool.