Aegis- AI-Powered SOC Platform

Aegis is an AI-powered Security Operations Center platform built on AMD Instinct GPUs. It combines forensic log analysis, real-time alert triage, continuous monitoring, and threat intelligence lookup into one tool for SOC analysts. Log Analyzer: Upload logs from five formats — Linux syslog, Windows Event logs, Apache/Nginx, firewall logs, and JSON. A 7-stage pipeline runs 15 detection rules with aggregate analysis (brute force, credential dumping, lateral movement, exfiltration), matches threat intel against known-bad IOCs with auto-escalation, maps findings to MITRE ATT&CK across 12 tactics as an interactive heatmap, generates Sigma/YARA/Splunk/Elastic detection rules per finding, runs LLM triage on top findings, and produces a PDF report. Analysts can chat with findings via natural language and export as CSV or JSON. Bulk multi-file upload enables cross-source correlation. Live Triage: Real-time alert queue processed by five specialized AI agents — Parser, Correlator, Triage, Planner, Reporter. Each has a narrow scope making the reasoning chain fully auditable. Every action requires human approval. Continuous Monitoring: Simulates a 24/7 SOC feed with a live dashboard tracking severity, latency, and IOC detection across cumulative runs. Threat Intel Lookup: Instant IP enrichment — IOC checks, CVE associations, ATT&CK mapping, and remediation recommendations. Fully offline, zero data leaves the environment. The platform runs on any OpenAI-compatible endpoint. Same code on AMD vLLM and Fireworks AI — zero code changes. Benchmarked on AMD Instinct showing throughput scaling from 293 to 506 tokens/sec when doubling concurrency with flat p95 latency, proving efficient GPU utilization for parallel security AI inference.