Final-year Integrated BS-MS student in Computational Sciences at IACS, Kolkata, India.
- 🔬 Researching provenance-based intrusion detection (PIDS) and GNN-based anomaly detection for APT detection
- 🛰️ Built PIDSMaker-viz, a GPU-accelerated tool for visualizing GNN embeddings in provenance graphs (more below 👇)
- 🎓 Applying to PhD programs in systems security & ML for graph-structured data
- ⚡ Off the clock: classic cinema, high-fidelity audio gear, and badminton
A GPU-accelerated visualization tool for inspecting GNN embeddings inside provenance-based IDS pipelines — built to make it easier to see why a model flags (or misses) malicious activity.
- ⚙️ Three-tier UMAP pipeline: RAPIDS cuML → batched GPU kNN (PyTorch) → CPU fallback
- 🖥️ Native Qt/VisPy desktop GUI with custom GLSL shaders for temporal animation
- 🔍 Epoch scrubbing, detection-status coloring, CSV-based node filtering
- 🔄 Word2Vec ↔ GNN encoder hot-swap, multi-hop coordinates across all 8 PIDSMaker architectures
🔗 github.com/pratyaynotfound/PIDSMaker-viz
Languages
ML & Data
Systems & Infra
Cloud & Other
Operating Systems
