Personal details

Kevin T. - Remote software architect

Kevin T.

Based in: 🇸🇻 El Salvador
Timezone: Central America (UTC-6)

About

AI Agent Engineer with 7+ years of enterprise Java Full Stack experience. Since December 2025 I have been building a production multi-agent AI fleet — starting with Project Hera — that now runs 24/7: autonomous agents (Cobalto, Hera, Scout) coordinating real client workflows via KVM computer control, pgvector semantic memory, WhatsApp pipelines, and the Claude/Anthropic API, all with zero human intervention. I work through an agent-orchestrated delivery model: I own architecture, specifications, code review, and QA gates while AI agents implement under my direction — shipping production systems at a pace no solo engineer can match by hand. That fleet is what earned me my current AI engineering engagements. Java background: 8.5 years at Applaudo Studios / Blackhawk Network delivering for Verizon Wireless, Comcast, Duke Energy, and Altice — 150+ pull requests across 20+ repositories, AWS Lift & Shift migrations of high-traffic US platforms, and daily production incident RCA. B2/C1 English; 8+ years of direct US stakeholder communication.

Work Experience

AI Agent Engineer & R&D Lead
AgentixVanguard | Jun 2026 - Present
Python
PostgreSQL
Docker
Chrome Extension
KVM
WhatsApp
WhyLabs
Pgvector

Lead a development team composed 100% of autonomous AI agents (PO, PM, Dev, QA) — acting as software architect and CI/CD lead for all agent workflows.

Delivered the entire platform portfolio (ExpressToGo, Hera 2.0, ERA 3.0, bridge infrastructure) through agent-orchestrated development — agents implement; I own architecture, code review, and QA gates.

Deployed Cobalto: 24/7 leader agent on local Windows infrastructure, coordinating tasks, memory, and sub-agent operations with no human relay.

Implemented KVM computer control via JetKVM + RTX 5080 GPU + Florence-2 OCR — enabling Claude to see and operate a full Windows desktop remotely in production.

Built pgvector semantic memory (Postgres + pgvector, 92+ synced context entries) shared across agents via custom bridge API — persistent cross-session context retrieval.

Architected WhatsApp→AI→systems pipeline: inbound messages trigger agent reasoning, tool execution, and real business actions autonomously.

Built Chrome extension bridge for DOM-aware browser automation: LinkedIn outreach, form completion, and web research fully autonomous.

Leading TFL (Agentix): AI voice widget platform — ElevenLabs 4-emotion voices, multi-tenant feature-flag widgets, RAG-JSON knowledge retrieval, AWS deployment (22 tasks, M0–M3).

CTO & Co-Founder
ExpressToGo | Apr 2026 - Present
Python
Docker
WhatsApp
Pgvector

Lead technical architecture for an AI-native delivery coordination platform powered by the Hera agent.

Hera handles inbound delivery orders → reasoning → real-time customer status updates via WhatsApp, fully autonomously.

Projects

Production AI Fleet
2025
Python
Docker
KVM
Claude
Pgvector
3-agent autonomous fleet (Cobalto · Scout · Hera) running 24/7. KVM computer control, pgvector memory, WhatsApp pipeline, Chrome browser automation. Built in parallel with full-time employment at Applaudo.
TFL Voice Platform
2026
Python
AWS
WhyLabs
Claude
Multi-tenant AI voice widget platform (Agentix): ElevenLabs 4-emotion voices, RAG-JSON knowledge retrieval, feature-flag widgets per tenant, AWS deployment.