Software & data engineer with 5+ years building production AI-powered agentic workflows, ML pipelines, and backend services. Currently shipping LLM and MCP-based tools at USAA driving multi-million-dollar monthly savings.
Near-real-time agentic workflow combining query telemetry, platform resource metrics, and multi-database context across UDB, Oracle, and SQL Server. MCP-style tool integrations expose structured performance context to LLM-driven automation, with a human-in-the-loop review model.
ML pipeline classifying millions of rows of query telemetry to surface inefficient patterns. Production tool delivering sustained $10K+/month cost reduction.
Secure, local-first AI agent platform orchestrating distributed nodes, agents, and tools via typed protocols, policy controls, and audit events. Full-stack control plane over WebSocket.
Self-hosted multi-agent AI environment extended with custom governance, service integrations, tool routing, gateway communication, and runtime observability.
Internal AI agent for structured automation, tool orchestration, and context-aware task execution across integrated services. Reusable agentic skill pattern.
Internal AI-powered knowledge platform using LightRAG to centralize tool governance, review status, use cases, cost visibility, and adoption guidance across teams.
Self-hosted multi-agent runtime with a custom CLI, integrated terminal dashboard, animated sprite UI, remote system controls, and hot-key privacy shortcuts. 12 specialized agents in a tiered cost-routing architecture.
Single source-of-truth for version-controlled skills, instructions, prompts, and rules — published to 8 agentic targets: Copilot, Claude Code, Codex, Cursor, Cline, BlackBox AI, Hermes, OpenClaw. 157+ skills and counting.
Layered guardrail system and memory consolidation pipeline for LLM agents — context compression with LeanCTX (80% reduction), structured long-term memory, schema-indexed Obsidian vault with 1,500+ entries, and policy-gated execution across the agent tier.
Active scraper fleet pulling AI, GitHub, and world economic news into the knowledge base in near real-time. Feeds downstream research agents and the trading pipeline's signal stage. Auto-downloads and security-scans trending agentic skills from open marketplaces.
Internal registry of 98 MCP servers exposing 500+ tools — governed, versioned, and auto-assigned to agents by profile and role. Integrated with orchestration routing and the tool registry hub for adoption tracking and cost visibility.
Built production ingestion and automation across 10+ enterprise data platforms — SQL Server, Oracle, Snowflake, UDB, Cassandra, Couchbase, Aurora/Postgres, Neptune, Netezza. Active Product Owner: scoping, prioritizing, and translating technical trade-offs to non-technical stakeholders.
Designed and shipped an ML-based classification workflow (XGBoost, DBSCAN, HDBSCAN, scikit-learn) identifying high-cost query patterns across multi-platform environments — generating $10K+/month in sustained savings. Now extending it with a near-real-time AI-assisted agentic recommendation layer.
Modernized enterprise SDLC by integrating Talon Batch automation, automated testing, and ticket-generation into deployment pipelines — improving consistency, release throughput, and operational traceability.
Contributing to n8n with hands-on exposure to workflow automation. Building local-first agent runtimes, RAG knowledge platforms, and reusable agentic skill patterns aligned with GTM automation.
Graduated with strong systems and security foundations.
No cloud platforms. No vendor lock-in. Everything lives on hardware I own and control. Twelve specialized agents in a tiered cost-routing architecture, a multi-stage trading pipeline, and a hardened infrastructure stack — all running 24/7 on local silicon.