Peru.N
§ Projects

Architecture, cloud & security case studies

Selected work covering system design, delivery decisions, security analysis, and implementation outcomes.

03 entriesSorted / recent
GraphRAG Incident-Intelligence Copilot architecture diagram
/ 01Architecture
Product · AI2026-07-05

GraphRAG Incident-Intelligence Copilot

Fused knowledge-graph structure with vector similarity so incident diagnosis is driven by evidence, not by a plausible-sounding guess.

Built an on-call copilot that fuses a service dependency graph with semantic search over past incidents, then has an LLM synthesize a grounded root-cause diagnosis and blast radius from ranked evidence.

AI SystemsPlatform EngineeringReliabilityKnowledge Graphs
  • Combined graph structure and historical precedent so root-cause candidates are both reachable and evidenced, not just semantically similar.
  • Reached a 1.0 hit-rate@3 and 0.944 MRR with hybrid retrieval versus 0.778 and 0.569 for graph-only in an ablation over chaos-injected incidents.
  • Turned incident labelling from an argument into a fact by generating ground truth through controlled fault injection.
View case studyRepository
Agentic Kubernetes IaC Security Reviewer architecture diagram
/ 02Architecture
Security2026-06-23

Agentic Kubernetes IaC Security Reviewer

Combined deterministic security scanning with bounded agent remediation to make infrastructure findings easier to trust and act on.

Built an agent-assisted CLI that turns Kubernetes, Terraform, and Dockerfile scanner output into prioritized security reviews, control-backed explanations, and safely staged remediation artifacts.

Security EngineeringPlatform EngineeringDevSecOpsAI Products
  • Turned raw infrastructure scan output into a prioritized review workflow engineers can act on faster.
  • Added control-backed explanations that make findings easier to defend in platform and security discussions.
  • Reduced remediation risk by staging generated fixes outside the source tree with bounded retry logic.
View case studyRepository
AI PDF Reader architecture diagram
/ 03Architecture
Product · AI2026-04-17

AI PDF Reader

A private, local-first document AI product with grounded retrieval, desktop packaging, and strong UX.

Built a local-first AI PDF Reader that lets users upload documents, retrieve grounded answers, generate study guides, and run as both a web app and offline desktop experience.

AI EngineeringSoftware ArchitectureLocal-First AISecurity Engineering
  • Delivered a product that can run in browser mode or as a desktop app for Windows and macOS.
  • Combined PDF parsing, chunking, embeddings, retrieval, and local LLM answers into one polished workflow.
  • Improved privacy posture by supporting local GGUF models and offline packaging for sensitive document use cases.
View case studyRepository