Edward Mwaura

Bachelor of Science in Computer Security and Forensics

Meru University of Science and Technology, Kenya

LLMShield — AI application security gateway

• Built a FastAPI gateway prototype with heuristic checks for prompt-injection patterns, PII, and selected secret formats.
• Added configurable, validated YAML policy rules with category-specific actions and severity thresholds, plus a mock provider, SQLite audit metadata, optional API-key checks, and a React dashboard.
• Added an authenticated operational metrics endpoint and dashboard summaries for recent audit decisions, finding categories, and severities, without exposing raw prompts or responses.
• Built a synthetic evaluation harness reporting precision, recall, false-positive rate, and false-negative rate separately for prompt-injection, PII, and secret detection.
• CI runs tests, linting, type checks, and the frontend build; evaluation results are documented as diagnostic synthetic metrics, not real-world benchmarks.

CloudSentry — Static configuration scanner

• Implemented 10 documented rule IDs covering selected Terraform, Dockerfile, and Docker Compose misconfigurations.
• Improved source-line attribution for Terraform resources and Compose services so findings point closer to the relevant configuration.
• Added text, JSON, and SARIF 2.1.0 reporting, including scan diagnostics and suppression justifications for compatible code-scanning workflows.
• Improved dashboard scan states and hardened reasoned suppressions, preserving suppressed findings in machine-readable output for audit.
• Made incomplete scans fail clearly rather than appear clean, with regression coverage for SARIF output, suppressions, source attribution, and exit-code precedence.