Secure software. Applied ML.
Immersive frontends.
A CS student from Coimbatore, obsessed with how systems break — and how good design makes them unbreakable.
Anomalies, spotted in real time.
A network anomaly detection system powered by Isolation Forest. Dark UI. Live logs. Sonic alerts. Built because security tools shouldn't feel like SharePoint.
Streaming anomaly scores update as traffic flows. Color-coded threats surface instantly, alongside per-host breakdowns.
View on GitHub↗Unsupervised. Lightweight. Trained on clean baselines — every outlier is its own story.
Threats trigger audible cues so operators don't need to stare at the screen.
No installer. No accounts. streamlit run aegis.py and you're monitoring.
A real shell. Click in and type help — or try about, projects, aegis run, matrix, theme. Tab completes. ↑/↓ history. ⌃L clears.
Every project is a live repository. Hover for the back, click to open source on GitHub.
Every public repo, wired by theme — like an Obsidian vault. Drag a node, hover to trace its links, click to jump straight to it on GitHub.
Not a LinkedIn-endorsed list. Only what's in my actual repos. Hover the cluster — they react.
Short notes from building the repos above — on how systems break, and what makes good ones hold.
Most breaches I read about aren't clever — they're a trusted input nobody re-validated. Now I write the assumption next to every boundary: who can reach this, and what happens when they lie?
Building Aegis taught me the model is the easy 20%. The other 80% is defining clean precisely enough that a real outlier can't slip in wearing a sloppy baseline as a disguise.
A 60fps interaction isn't decoration — it's trust. If the page stutters, the user quietly decides the system is fragile. I budget motion the same way I budget memory.
Happy to chat about internships, collaborations, or that one bug you can't reproduce at 2 AM.