Building multi-agent AI systems

Adolfo Lugo

Senior AI Engineer

Jacksonville, FL · Remote

I build multi-agent AI systems, RAG pipelines and the cloud infrastructure they run on.

About

Secure, observable, recoverable AI

Building AI systems isn't just about calling a model, it's about creating agents that remember what they learned, check their own work, and hand off cleanly to each other and to the people they serve.

I'm a Senior AI Engineer with 10+ years in enterprise IT, Windows and Linux systems administration, and cloud infrastructure. For the past 3 years I've focused on production RAG pipelines, multi-agent LLM workflows, and fast web apps. I came up through Tier 2 support, server administration, and on-call rotations, so I build AI the way I ran servers: secure by default, observable, and recoverable.

Enterprise RAG

Ingestion, semantic chunking, embeddings, hybrid retrieval, and strict source grounding.

Agent engineering

Structured tool calling, MCP connectors, persistent agent memory, eval harnesses, and human-in-the-loop guardrails.

Cloud & DevOps

AWS, GitHub Actions, CodePipeline, Terraform, Docker, Vercel, and Supabase.

Infrastructure & security

Windows Server, Active Directory, RHEL/Ubuntu, VMware, DISA STIG, CIS, NIST 800-171, SOX, and PCI.

Projects

What I'm building

Live products link out. Private builds run behind sign-in or on internal infrastructure, so they're described here instead.

Private build

Gemini Agent Team

Multi-bot AI platform where Gemini bots create other bots, delegate work, and collaborate on shared tasks, with a web app for chats, projects, team, and usage.

  • Supabase Edge Functions (Deno/TypeScript)
  • Gemini API (function calling, model fallback chain)
  • OpenAI / xAI / Meta Llama provider support
  • Supabase Postgres
  • pg_cron + pg_net worker
  • Supabase Auth (per-bot logins)
  • Vercel (static web app)
  • HTML/CSS/JavaScript
Gemini Agent Team chat screen on mobileBot designer and instructions editorUsage meter and model fallback chain settings
Private build

Agent Hub

Shared memory and task queue that lets Gemini and Grok agents read each other's knowledge, claim tasks, and log events safely.

  • Supabase Postgres
  • pgvector (768-dim, HNSW)
  • Row-Level Security per agent
  • Postgres RPCs (match_facts, claim_task, complete_task)
  • pg_cron
  • Gemini embeddings
  • Python
  • Google Drive API (daily Markdown mirror)
Live

Tasks for Humanity

Mobile-first citizen-science platform that breaks open research problems into small, verifiable tasks anyone can do with their own AI. Hybrid vector + full-text search and row-level security on every table.

  • Next.js
  • TypeScript
  • Tailwind CSS
  • Supabase (Postgres, pgvector, tsvector hybrid search, Auth, RLS, RPCs, triggers)
  • Vercel + Vercel Cron
  • Open Targets GraphQL
  • ClinicalTrials.gov API
  • OpenAlex
  • PubMed
Live

UpIsUp Multi-Agent Operations Platform

Specialized AI agents that run research, analytics, content, and operations for the UpIsUp learning community on scheduled routines.

  • Python (APScheduler, systemd)
  • MCP tool connectors
  • Evaluation/grading harnesses
  • JSONL cost and usage ledgers
  • Discord
  • Whop
Live

Foundational IT

IT and AI services site; part of a Website-as-a-Service automation engine for local businesses.

  • Next.js
  • TypeScript
  • Supabase
  • Python site-generation pipeline
  • Webhook intake
  • Vercel edge deployments
Live

Scale with FITOS

Website-as-a-Service platform that takes local businesses online with generated, automatically deployed sites.

  • Next.js
  • TypeScript
  • Supabase (Postgres + pgvector)
  • Python site-generation pipeline
  • Webhook intake
  • Vercel edge deployments
Prototype

Skills (AI-agent marketplace app)

iOS/Android marketplace and feed for single-purpose AI agents, with a skill compiler that adds schema guards, LLM-as-judge checks, and self-healing retries.

  • Expo SDK
  • React Native
  • expo-router
  • TypeScript
  • Zustand + AsyncStorage
  • Reanimated
  • EAS Build
  • Supabase/Next.js backend (designed)
Private build

Cost-Capped Agent Job Runtime

Moved recurring agent jobs from LLM routines to deterministic Python with kill switches, dry-run sandboxing, and daily LLM budget caps.

  • Python
  • APScheduler
  • systemd/cron
  • Linux mount namespaces/overlayfs
  • RouteLLM
  • JSONL ledgers

Experience

10+ years from Tier 2 support to AI engineering

  1. Senior AI Engineer

    Navient
    Jan 2024 – Present · Remote / Herndon, VA
    • Partnered with business and product teams to modernize enterprise data pipelines: automated document ingestion, semantic chunking, pgvector embeddings (1536-dim), hybrid retrieval (tsvector + vector), and strict source grounding, deploying IAM- and RLS-governed AI agents on AWS.
    • Built grounded enterprise RAG over Confluence runbooks, Jira ticket histories, and PostgreSQL operational records, with Python evaluation harnesses (pytest, LLM-as-a-judge logic guards, self-healing validation loops) to reduce hallucinations.
    • Engineered full-stack agentic apps with Next.js, React, TypeScript, REST/GraphQL/WebSocket APIs, and AWS, including short- and long-term agent memory, Model Context Protocol (MCP) tool connectors, and multi-hop RAG over Pinecone and PostgreSQL.
    • Built deterministic LLM workflows in TypeScript, Node.js, and Python with strict JSON-schema tool calling across Ingestion, Reasoning, and Verifier agents, human-in-the-loop approval gates, and streaming web interfaces.
    • Migrated scheduled agent routines to deterministic Python (APScheduler/systemd) with mount-namespace dry-run harnesses, RouteLLM model routing, daily token budget caps, and JSONL usage ledgers.
    • Built cross-account CI/CD with GitHub Actions, CodePipeline, CodeBuild, CodeDeploy, and Terraform, and moved a compromised legacy Joomla site to a secure S3 + CloudFront static/edge architecture.
    • Supported containerized Python/Node.js AI services on Docker, administered hybrid Windows and RHEL/Ubuntu systems across VMware and AWS, remediated WorkSpaces vulnerabilities, and enforced SOX, PCI, and NIST 800-171 controls with Active Directory, SSSD, and RBAC.
  2. Windows & VMware Systems Administrator

    Rotech
    Mar 2023 – Jan 2024 · Orlando, FL
    • Administered hybrid Windows Server and Linux infrastructure on VMware ESXi 7.0 hosts, maintained VM templates, managed Active Directory Group Policy, and troubleshot VMware Horizon Cloud Connector issues.
    • Wrote PowerShell and Ansible playbooks for server hardening, user provisioning, BigFix removal, and routine admin automation, reducing manual ticket handling.
    • Hardened servers against DISA STIG and CIS benchmarks and set up systemd/service recovery scripts so applications came back automatically after monthly patching.
    • Deployed secure SFTP and SMB/Samba file transfer services, tuned performance, and resolved DNS, DHCP, and network connectivity issues.
  3. Windows & Infrastructure Systems Administrator

    Highmark
    Sep 2017 – Mar 2023 · Remote / Pittsburgh, PA
    • Built a PowerShell/Bash utility that updates and recovers Oracle JDBC connection strings and app configurations, turning a long manual process into a near-instant one across multiple applications and environments.
    • Managed the full lifecycle of Windows Server and RHEL systems across VMware vSphere/vCenter, bare metal, and AWS: builds, Active Directory, WSUS/SCCM patching, upgrades, backups, and performance tuning.
    • Automated server configuration and repetitive admin work with PowerShell, Ansible Tower, Bash, Git, and GitHub.
    • Configured Active Directory, Group Policy, file permissions, DNS, LDAP, SSL certificates, firewalls, and SFTP.
    • Monitored infrastructure with Dynatrace and vRealize Operations, analyzed logs, and took part in production on-call and incident response.
  4. Systems Support Analyst / Tier 2 Help Desk Specialist

    Stoltenberg Consulting
    Jan 2015 – Sep 2016 · Remote / Bethel Park, PA
    • Provided Tier 2 remote support for Windows workstations, shared drives, OneDrive, Citrix, Cisco AnyConnect VPN, and Microsoft Office 365.
    • Diagnosed hardware, software, identity, and network issues with PowerShell/CMD, RDP, SCCM, and remote-support tools.
    • Managed incidents in ServiceNow within strict SLAs, documented fixes in the knowledge base, and escalated complex issues to the right teams.
    • Supported infrastructure upgrades (Dell servers, Cisco routers and switches), Active Directory user administration, and Azure MFA setup.

Skills

Stack

AI, RAG & Agents

  • Multi-Agent Orchestration
  • Retrieval-Augmented Generation
  • Model Context Protocol (MCP)
  • LLM Tool / Function Calling
  • Persistent Agent Memory
  • LLM Evaluation (LLM-as-a-Judge)
  • Human-in-the-Loop Guardrails
  • RouteLLM
  • Gemini
  • OpenAI
  • Anthropic Claude
  • Grok

Data & Vector

  • Supabase
  • PostgreSQL
  • pgvector
  • Pinecone
  • Hybrid Search (vector + tsvector)
  • Row-Level Security
  • SECURITY DEFINER RPCs
  • pg_cron

Full-Stack

  • Python
  • TypeScript
  • JavaScript
  • Node.js
  • Deno
  • React
  • Next.js
  • React Native (Expo)
  • Tailwind CSS
  • REST
  • GraphQL
  • WebSockets

Cloud & DevOps

  • AWS (EC2, S3, CloudFront, IAM, CloudWatch)
  • Vercel
  • Supabase Edge Functions
  • GitHub Actions
  • CodePipeline / CodeBuild / CodeDeploy
  • Terraform
  • Ansible
  • Docker
  • systemd / APScheduler

Infrastructure & Security

  • Windows Server
  • Active Directory / Entra ID
  • Group Policy
  • RHEL / Ubuntu
  • VMware vSphere / Horizon
  • PowerShell
  • Bash
  • DISA STIG
  • CIS
  • NIST 800-171
  • SOX / PCI
  • Splunk
  • Dynatrace
  • ServiceNow

Contact

Let's build something

Open to teams building agentic AI, LLM platforms, and cloud infrastructure.