Forward Deployed Engineer
Oct 2025 — Present
PwCTampa, FL · Hybrid
Multi-agent threat-intelligence platform — Built a production system that autonomously maps an organization's external attack surface, replacing manual reconnaissance with an LLM-augmented agent swarm.
Architected a microservices system with a FastAPI supervisor orchestrating containerized worker agents over RabbitMQ, using PostgreSQL 16 + pgvector for state and Redis for caching, pub/sub, and rate limiting.
Engineered a deterministic-tools-first orchestration engine that constrains LLM output to strict JSON schemas with a JSON-repair layer and Langfuse tracing/evals — eliminating hallucinated intelligence in a security-critical domain; added a resilient dispatch layer so a broker outage degrades gracefully instead of hard-failing.
Deployed to Kubernetes via Argo CD GitOps and Helm with a GitHub Actions pipeline building 7 service images in parallel across dev/qa/stage/prod; secured with Azure Entra ID, Key Vault, and server-side RBAC.
AI-assisted risk-assessment microservice (FastAPI + MongoDB) grounding control generation in the NIST 800-53 Rev 5 catalog via a deterministic-retrieval → LLM-judgment → human-in-the-loop pipeline.
Built a multimodal evidence-review pipeline (vision models, openpyxl, pypdf) that rates control effectiveness, backed by an offline evaluation harness; exposed the full REST surface as MCP tools for tool-calling agents.
Streamed live pipeline progress to a Vue 3 / SSE frontend for real-time analyst visibility.
