FastAPI · Next.js · Docker · Docker Compose · AWS EC2 · Nginx · GitHub Actions
LogSense is a full-stack app I built with my team to catch weird behavior in system logs before it turns into a real problem. It watches a stream of logs, tracks things like error rate and batch size over a rolling window, and uses a Z-score based statistical engine to spot when something looks off. No machine learning model, no training data needed. It just learns what normal looks like and flags what doesn't fit.
On top of the statistics sits a rule-based AI agent that decides how serious an anomaly actually is. It weighs different signals against each other, an empty batch, an all-errors batch, a high or medium Z-score, and gives back a plain-language reason for its verdict: normal, warning, or critical.
The backend is FastAPI, the dashboard is Next.js and polls for new alerts every 5 seconds so you can watch things update live. The whole thing runs on Docker and Docker Compose, deployed to AWS EC2 with Nginx and HTTPS, shipped through a GitHub Actions CI/CD pipeline.
I owned deployment and infrastructure on this one. My teammate Ammar Hussain built the backend and AI agent logic, and Hassaan worked on the CI/CD side with me. We also benchmarked LogSense against a plain fixed-threshold detector using the public HDFS log dataset, to prove it actually catches more with fewer false alarms, not just that it runs.
Demo