Cytex.io

The problem

Security teams drown in raw threat telemetry. The signal that matters arrives mixed with noise, and analysts spend their time triaging alerts instead of responding to the ones that are real.

The approach

Built an AI-powered cyber threat intelligence platform with LLM-driven reasoning modules that correlate and prioritise incoming signals. Python and LangGraph orchestrate the analysis agents, FastAPI serves the API layer, Next.js drives the analyst-facing console, and Postgres on AWS holds the intelligence store.

Results

• Near real-time threat analysis instead of batch triage.
• LLM reasoning modules surface prioritised findings rather than raw alerts.
• Analyst console built on Next.js for live investigation workflows.

Engineering notes

LangGraph gives each reasoning module explicit, typed state, which keeps agent behaviour auditable after the fact — a requirement in a security context.