Projects
I build small, practical tools to understand how data, assumptions and models change a decision. The projects below range from financial analysis to research workflows and applied machine learning.
Financial Data & Scenario Tools
A browser-based toolkit covering live FX and market data, country inflation, DCF valuation, investment returns, compound interest, loan amortisation and purchasing-power analysis.
The DCF tool makes its assumptions visible and adjustable, so users can see how cash flow, growth and discount rates change the valuation rather than treating the output as a black box.
Why I built it: I learn finance more effectively when I can test an assumption, see the result immediately and understand which input actually drives the answer.
DCFScenario analysisJavaScriptWorld Bank APIMarket data APIs
ASML Valuation Tracker
A live tracker that compares ASML's market price with an independently modelled DCF range. It combines company research, explicit assumptions and scenario analysis in one compact view.
What it taught me: valuation is less about producing one precise number than understanding which expectations are already reflected in the price and where the downside assumptions break.
Equity researchDCFRisk scenarios
Video Annotation Tool
A lightweight browser-based video labelling workflow with frame navigation, timestamped multi-class annotations and CSV export. It was designed for a research task that needed speed and simplicity rather than a large annotation platform.
Design principle: start with the user's real bottleneck. The most useful version had fewer features, a shorter workflow and clearer output.
PythonFlaskOpenCVJavaScript
Offline Music Recommendation
An offline pipeline combining speech recognition, sentiment analysis, content-based filtering and local playback. The project explored how separate models can be connected into one usable workflow under strict privacy and connectivity constraints.
PythonSpeech recognitionSentiment analysis
csvtool
A C++ CSV parser built when Python-based processing became the bottleneck for a large research file. The project focused on runtime, memory use and a simple interface rather than adding unnecessary abstraction.
C++PerformanceData processing
AI-Eco
An interactive visualisation of AI's energy, water and e-waste footprint, originally developed as a university interaction-design project. Users can compare models and explore how usage assumptions change environmental impact.
Three.jsChart.jsInteractive storytelling