Ricardo Ledan

• Built a new AI capability within Deloitte's internal AI startup that landed a federal intelligence agency as a client. Recruited to R&D for a unique blend of fullstack engineering and AI/ML skills. Designed and pitched a multi-agent knowledge base for consuming SARs reports, automating analysis that previously required multiple analysts over days.
• Owned end-to-end system architecture for a production agentic AI platform serving a federal intelligence agency — from inference pipeline design and vector store configuration to deployment on AWS GovCloud. Built the foundational application template and led a 3-person team (data scientist + 2 AI engineers) from zero to production.
• Engineered a GraphRAG pipeline that discovered 67% more entities than traditional RAG through multi-hop graph traversal. Four-stage architecture: document ingestion, YAML-driven knowledge extraction via GPT-4, Neo4j graph analytics (centrality, community detection, entity resolution), and hybrid retrieval combining vector search with graph traversal.

• Removed the biggest bottleneck in a top-tier bank's credit card merchant onboarding pipeline, where manual regex rule creation was capped at 226 rules/month. Led the integration of LLM capability into the rules engine, projected to reduce manual effort by 70–85%.
• Designed and presented a multi-option technical roadmap to client executives, evaluating regex enhancement, supervised ML, LLM integration with human-in-the-loop feedback, and a long-term unified merchant identifier strategy. The recommended LLM approach was approved for implementation.
• Shipped the production-ready application in three weeks — rewrote the Python codebase to internal quality standards, added test coverage, built CLI tooling for the rules team, and packaged the application for local distribution across team machines. Coordinated delivery across onshore and offshore teams.

• Contributed to a federal WMD detection system on AWS GovCloud, building infrastructure across a classified environment. Designed CI/CD pipelines with end-to-end encryption, RBAC, and NLP processing workflows using AWS Comprehend.
• Built TrustGen, a blockchain proof of concept for data provenance and access control that gained sponsorship from a Deloitte partner. Implemented on-chain dataset tracking and secured data retrieval on Hyperledger Besu.

Led infrastructure engineering for a multi-cloud analytics platform (AWS EKS, Azure, GCP) — designing Kubernetes-based deployment pipelines, Terraform IaC modules, and CI/CD standards adopted org-wide. Mentored 3 engineers and established trunk-based development and automated test coverage requirements.