Skip to content
Selected work

Supply-chain security

From thousands of findings to the few that matter

An SBOM-driven vulnerability intelligence system that reduces raw scanner output to actionable risk with contextual validation and LLM-assisted prioritisation.

AI Platform Security Engineer · ISS STOXX · details generalised; specifics stay internal

The problem

Dependency scanners report everything that matches a version range. Across hundreds of repositories that becomes thousands of findings — and teams stop reading them.

The approach

  1. 01Built the system on SBOMs across hundreds of repositories, so every answer starts from what is actually shipped.
  2. 02Validated each finding in context before it reached a team, so only real exposure made the list.
  3. 03Used LLM-assisted prioritisation to rank what remained, with a human-readable rationale for each risk.
  4. 04Used the same inventory to answer “who is affected?” within hours of a new open-source disclosure.

The outcome

Thousands of raw findings reduced to a few hundred actionable risks, and every project affected by newly disclosed open-source vulnerabilities identified for remediation.

Stack

SBOM · SCA · OSV · Threat validation · LLM-assisted triage