Full-Stack AI Engineer & Security Researcher
Crafting intelligent, secure, and high-performance systems across deep learning, computer vision, and proactive threat intelligence.
| 🎯 Focus | 🛠 Stack | 🚀 What I Build |
|---|---|---|
| AI / ML | PyTorch · TensorFlow · HuggingFace · Scikit-Learn | Vision systems, predictive models, automation pipelines |
| Backend | FastAPI · Node.js · Express · Go-Fiber | APIs, secure services, real-time data flows |
| Frontend / Mobile | React · Next.js · Android | Clean interfaces, dashboards, mobile experiences |
| Infra | Docker · Kubernetes · AWS · PostgreSQL · Redis · GitHub Actions | Scalable deployments, CI/CD, observability |
PRISM is a proactive threat intelligence and security risk assessment platform that helps identify, analyze, and mitigate cyber threats before they impact infrastructure.
- Proactive threat hunting with automated anomaly detection and traffic classification.
- Real-time risk scoring for faster incident prioritization.
- Interactive dashboards for telemetry, triage, and response workflows.
Risk tiers
- Critical: active exfiltration indicators or anomalous SSH behavior.
- High: repeated failed logins combined with suspicious origin signals.
- Medium: internal scanning or configuration drift.
Quick self-hosting
git clone https://github.com/akilaisadev/prism.git
cd prism
make setup
make up
An end-to-end semantic segmentation pipeline built with U-Net and a ResNet50 backbone to identify informal settlements in high-resolution satellite imagery.
Validation metrics
| Metric | Score |
|---|---|
| AUC-ROC | 0.942 |
| F1-Score | 0.897 |
| IoU (Slum Class) | 0.814 |
| Precision | 0.912 |
| Recall | 0.883 |
Inference example
import torch
from slum_detector import SlumDetectorUNet
detector = SlumDetectorUNet.load_from_checkpoint("best_model.ckpt")
detector.eval().to("cuda" if torch.cuda.is_available() else "cpu")
satellite_image = detector.preprocess_image("data/tile_0912.png")
with torch.no_grad():
prediction_mask = detector(satellite_image)
detector.save_overlay(prediction_mask, "output/slum_overlay.png")
print("Inference completed. Overlay saved to output/slum_overlay.png.")
Made with ❤️ by Akila Wasalathilaka