AIVSS
An AI-driven vulnerability scoring system
- Year
- 2025
- Role
- Systems + ML Engineer
Overview
AIVSS scores vulnerabilities using semantic embeddings and a tuned C++ scoring pipeline.
Evaluates security risks using embeddings and low-latency C++ pipelines backed by RocksDB and uSearch.
Problem
Security teams struggle to prioritize vulnerabilities consistently — rule-based scoring misses context and semantic similarity across large, noisy datasets.
Approach
Built an embedding-driven scoring pipeline in C++ with RocksDB persistence and uSearch indexing so risk signals stay fast, explainable, and low-latency at scale.
Highlights
- Semantic vulnerability scoring with vector embeddings
- Low-latency C++ pipeline tuned for production workloads
- Persistent storage and indexing for repeatable audits
- Red-team tooling integration for validation loops
Stack
C++PythonRocksDBuSearchEmbeddings