Research · 2026
AMLynx
Our research project on spotting money laundering as the transactions come in. A simulator streams them over ZeroMQ, and a scoring engine checks each one with rules and graph analysis, plus an optional ML filter. An analyst follows the results on a desktop dashboard. I built the algorithm and the software, my team handled the documentation, and we presented it as a paper at ICT4SD 2026.
- It flagged too many false positives at one point, so I spent 7 days and nights fine-tuning it
- Handles 400+ transactions a second with under 2ms latency when the system isn't under load
- The audit log is chained with HMAC, so you can tell if someone edited it
- Each alert lists the rules that fired
- The simulator knows the right answers, so you can score the engine while it runs
Python, PyQt6, ZeroMQ



Running on simulated transactions.