$ whoami▊
I build multi-agent AI systems, MCP-powered tooling and cloud-native backends. 8+ years shipping software across health, finance, telecom & logistics — now focused on production-grade LLM deployment patterns across AWS, Azure, GCP & Cloudflare.
¡Pura vida! 👋 I'm Jeremy — a software engineer based in San José, Costa Rica. What started as a university passion became a career dedicated to building innovative, scalable solutions.
I've had the privilege of working across health, finance, telecom and logistics. Today my focus is squarely on AI agents, multi-agent architectures and cloud-native platforms — designing systems where LLMs discover their own tools via MCP instead of hardcoding them.
I also led digital literacy initiatives at PANTEC (TEC, 2017–2019), bringing technology inclusion to underserved communities.
const developer = {
name: 'Jeremy Live',
role: 'AI Engineer',
based: 'San José, CR 🇨🇷',
stack: ['Python', 'TS', '.NET'],
focus: ['Agents', 'MCP', 'RAG'],
cloud: ['AWS', 'GCP', 'Azure',
'Cloudflare'],
status: 'Building the future'
};
I'm a software engineer — I don't do motivational talks. I apply the same system-architecture mindset I use to ship production systems to how teams run: management, communication, strategy, sales and culture. A 6-week intensive that turns leadership into a measurable, repeatable playbook.
Week 1 Leadership Audit, weeks 2–3 Communication & Culture Framework (huddles, 1:1s, decision-making, conflict de-escalation), weeks 4–6 Strategy & Sales Alignment.
No cheerleading. Repeatable frameworks and a scalable playbook your team actually runs — built for founders, self, small business (<50) and large orgs.
50% upfront, 50% at week 3. Optional monthly advisory retainer at $1,200/mo to keep the system running long after the intensive ends.
Build a complete production agentic software product — not an MVP — using four open-source agentic projects I use every day to ship GenAI apps at etheroi.com: opencode, openclaw-etheroi, hermes-agent-etheroi and openclaude. Run all models on API keys with a daily free tier.
The open-source AI coding agent — build & plan agents, LSP, MCP, 20+ providers.
github ↗Personal AI assistant — your product's assistant across 20+ channels, voice & canvas.
github ↗Self-improving agent — a built-in learning loop, memory & subagent delegation.
github ↗Coding-agent CLI router — one workflow across 20+ cloud & local model providers.
github ↗Agentic fundamentals, all four projects, system design, production hardening, deployment & a real capstone — your client's agentic product, shipped in production.
All models via API keys from providers with a daily free tier — OpenCode models, Google Gemini & OpenRouter. Build and launch without paying for tokens.
Build and ship a complete, playable video game — in production — by orchestrating multiple open-source AI agents that generate gameplay, build worlds, write narrative, produce assets and test balance alongside you. Idea to finished game, no concept-stage graveyard.
I'm comfortable across Unity 4 & 5, Maya + Python and open-source agentic projects — and every model runs on API keys with a free daily tier.
Scope your idea into an AI-aware Game Design Document and stress-test it with simulated players.
RAG-powered lore, dialogue, quests and levels that stay consistent and adapt to the player.
Keep every asset on-brief — preserving your game's aesthetic while scaling production.
Autonomous playtesters (bug reports, balance sims) so data drives the final polish.
Foundations, design, agent crews, and content pipelines, level generation, narrative engines, adversarial QA and a production pipeline — ending in your playable capstone.
All models via API keys from providers with a daily free tier — OpenCode models, Google Gemini & OpenRouter. Build a playable game before paying a cent.
Original AI products hosted on Cloudflare (R2 / D1 / KV / Workers AI / AI Gateway). Specialized in Generative AI architecture & multi-agent systems, with advanced deployment patterns for LLMs across AWS, Azure, GCP and Cloudflare.
Built three connected financial apps: an admin/client portal duo in PHP/Laravel, plus a Python app with Gemini AI & voice agents talking to a Cloudflare D1 relational database.
Advanced web solutions with agile practices; integrated AIOps features into KhayaCore. Scalable REST APIs, microservices with modern design patterns, CI/CD with Docker.
Real-time delivery platform on GCP with connected admin & driver portals, route optimization and public client forms. −30% order processing time · +25% driver productivity · −20% operating costs
Enterprise .NET 8 MVC app with public site + admin portal. Full Java API integration with business logic for client-facing forms. Projects included an insurance quoting system and a SHELL self-service system.
SaaS platform with MariaDB, Node.js, React/Redux & TypeScript. Full payment system with Stripe + PayPal subscriptions & webhooks, GA4 analytics, multi-language/multi-region support, and Artillery stress testing.
Designed and shipped 37 ETL pipelines with creative solutions — complex SQL Server stored procedures, C# scripting inside SSIS and PowerShell automation.
Responsive websites with improved design workflow; new features on the client's maintenance platform while juggling multiple concurrent tasks.
High-performance web apps with PHP/Laravel and Java/Spring; RESTful APIs in Python/Flask with PostgreSQL/MySQL/MongoDB. Responsive interfaces with Angular & Bootstrap, OAuth 2.0 / JWT auth, CI/CD with Docker & Kubernetes.
Instituto Tecnológico de Costa Rica (TEC) · Cartago
Focus on algorithm analysis, artificial intelligence, machine learning and mathematics.
Intl. Collegiate Programming Contest
Competed representing TEC — advanced algorithms, data structures and problem-solving under pressure on a world-class team.
Digital Literacy Program
Led digital literacy initiatives reaching 500+ people in underserved communities across Costa Rica — tech workshops and digital-inclusion projects with educational & governmental institutions.
Production AI products from the etheroi.com ecosystem — all live, all built end-to-end.
Multi-Agent-as-a-Service platform with LangChain — persistent chat sessions, workspaces, gallery storage in D1/R2, RAG deep-research across Azure, GCP & AWS.
Model-agnostic LangChain/LangGraph agents powered entirely by MCP tools over HTTP/SSE. Zero manual tool wiring — agents discover and call tools dynamically.
Multi-page AI platform — the hub of a growing ecosystem of one-shot AI webapps: barber, dentist, seller, clean, tattoo, paynow, B2B WhatsApp and more.
Voice-first AI application powered by Gemini — natural spoken interaction with intelligent agents.
Multi-language AI translator — fast, contextual translation as a lightweight edge-hosted webapp.
Fintech payment system with AI-generated reporting, connected to loan admin & client invoice portals for the Costa Rican market.
Autonomous agent chat platform powered by multiple AI models — multi-modal assistants with agentic capabilities.
Next-gen NFT platform — create, collect and trade unique digital art with auctions, artist galleries and blockchain-backed ownership.
gaja-dev · contnowtax · app-b2b · ecommerce-ios · estudiobelleza …
An active playground of demos, agents and models — from in-browser WebGPU agents to image & video generation.
In-browser coding agent — Soyuz 4B running fully client-side via llama.cpp WebGPU.
● running on zeroGemma 4 inference directly in the browser with WebGPU acceleration — no server needed.
★ 4 likes · staticA small LLM writes, draws & directs original ASCII films — generative storytelling experiment.
pinned spaceOpen-source cybersecurity agent — autonomous security analysis in a Gradio interface.
pinned spaceCoordinated agent swarm performing deep research tasks with shared state.
gradioLightning-fast image generation demo — part of the 2026 GPU demo collection.
pinned spaceLeadership, agentic AI and full game development — each one a system, not a talk. Tap a banner to open the full program.
Open to senior engineering roles, AI architecture consulting and ambitious collaborations — anywhere in the world, remotely.