BASIS SDK
Behavioural anomaly detection shipped as Django middleware, with the experiment suite used to prove it works.
- Python
- Django
- scikit-learn
- Isolation Forest
- One-Class SVM
- +2
cd ~/projects
Things I built to answer a question, with the reasoning left in. Every entry has a case study covering the problem, the approach, and where it falls short.
9 projects shown.
Behavioural anomaly detection shipped as Django middleware, with the experiment suite used to prove it works.
Static behavioural analysis, MITRE ATT&CK mapping and an LLM reasoning layer — inferring intent without ever running the file.
Banking fraud modelled as structure rather than rows — laundering loops and shared-ownership rings surfaced in Neo4j.
A self-hosted workspace — notes, files, scraping and an assistant running on a local model rather than someone else's API.
A long-form writing environment where the persona, the reader profile and the whole conversation are part of the prompt.
A Discord bot with no API key — local language models answering in a channel, with an editable persona and resettable memory.
The Hypixel SkyBlock mod the username came from — fourteen modules of dungeon tooling, rendering and automation.
A browser aim trainer with a hand-written game loop — pointer lock, canvas rendering and per-session reaction analytics.
Three trained scikit-learn models behind a health recommendation dashboard — regression for calories, classifiers for plans.