Jordi Lluis
I build, test and debug software to solve everyday problems. I use AI tools throughout development and explore new technologies to understand where they help—and where they fall short.
Research tools · recorded case
Which claims are actually supported by the sources? Compare passages, inspect disagreements and review a draft before saving it. The public case is recorded; live research runs in a separate local workspace.
Source-linked findings · human approval · retrieval evaluation
Interactive demo · market data
Choose a market, inspect a candle pattern and rewind the chart. Indicators are recalculated using only candles available at the replay cursor, so later prices cannot leak into the reading.
Read the walkthrough
Time-boundary tests · validated candle API
PostgreSQL runs locally and in CI, not in the public demo.
Operated system · historical case
A trading bot that reacted to order-book prices. I operated it on AWS, investigated execution and account discrepancies, and published the postmortem when the opportunity stopped working.
Read the engineering case
WebSocket books · signed orders · account reconciliation
Live trading stopped. Not proof of profitability.
Public energy data
Compare electricity prices with the power generated at the same time. Missing and delayed readings stay visible, so the comparison does not hide gaps or mix different periods.
Read the walkthrough
Generation: 33 countries · Prices: 21 zones
European coverage · no forecast claim.
My earlier work involved furniture dispatch, airport cargo records, route planning and futures trading. Those experiences gave me practical problems to investigate in software: mismatched records, changing information and decisions made under time constraints.
Experience, education & downloadable CVs ↗Fly Brain — testing whether a published fruit-fly connection map helps a simulation carry a signal better than random wiring. Under fair, written-in-advance comparisons on two groups of neurons, it did not; the next question is memory.