Production Sprint
One stuck AI feature, taken to production-ready on a fixed scope and a flat fee. You walk away with a shipped feature, an evaluation harness so you can trust it, and a runbook your team can own.
Booking 2 client slots for Q3 2026
I ship your AI features into production.
Your AI demo wowed the room — then stalled. Too slow, it hallucinates, it costs too much, or nobody trusts it enough to ship. I'm the forward-deployed engineer who closes that last mile: reliable, evaluated, shipped. 14+ years building at scale; AI live in production at ImmuneFi.
Shipped production software for
ImmuneFi
Thrive Market
Modus Create
Hobsons · Naviance
How I help
Three ways to work together
Start small to de-risk, then scale into an embedded partnership.
Embedded AI Engineer
A senior, forward-deployed engineer inside your team — shipping AI features continuously and owning the AI roadmap. The fastest way to turn "we should use AI" into things your customers actually use.
AI Architecture & Review
For teams that have the engineers but need direction. Architecture reviews, evaluation strategy, model and cost decisions, and a weekly call to keep your AI work on the rails.
Proof
Autonomous AI I've shipped
Not demos — live systems that take real-world actions.
Fiscalify
fiscalify.com ↗
An agent that reads a receipt photo (Claude vision), then drives Playwright browser automation across 18+ real merchant portals — OXXO, Walmart, Costco — to generate government-compliant CFDI invoices with no human in the loop. Mobile app, dashboard, and agent backend.
18+
portals automated
0
humans in the loop
Insite
client work
Document intelligence for construction teams: ask questions across RFIs, submittals, and change orders and get cited answers. Claude Agent SDK plus a custom 9-tool MCP server, multi-tenant, fully observable.
Rooxai
rooxai.com ↗
An agent that sells from your website and WhatsApp — not a chatbot. Conversational sales with payments built in.
ImmuneFi
2022 – 2025
AI products for the largest Web3 bug-bounty platform: Guardian AI (40% efficiency gain), a spam & duplicate classifier at +90% accuracy / 70% less manual review, and internal RAG systems — all live in production.
AgentFounder
agentfounder.ai ↗
An autonomous founder that builds and grows products — multi-agent orchestration in a native desktop app. Also explore contix.mx and midastools.com.
About
14+ years shipping software at scale — now fully focused on getting other teams' AI to production.
Independent AI Consultant — armando.mx
Forward-deployed AI engineering for startups and enterprises: stalled LLM features taken to production, agentic systems, RAG pipelines, and AI architecture advisory.
AI Engineer — ImmuneFi
Shipped Guardian AI, Hackoon, a spam/duplicate classifier (+90% accuracy), and internal RAG systems to production at the largest Web3 bug-bounty platform.
Engineering Manager — Thrive Market
Led a 12-engineer web team on a high-traffic e-commerce platform. 25% performance improvement, 40% fewer deployment errors.
Software Engineer — Thrive Market · Modus Create · TopBox
Next.js migrations, microservices, component libraries, analytics platforms. Co-author of Learning Ext JS 4 (Packt). B.S. Computer Science, UTM.
Questions
What you might be wondering
A forward-deployed AI engineer embeds directly in your team to make AI actually work in production — not just prototype it. I take an LLM feature that's stuck at the demo stage and make it reliable, evaluated, fast, and shipped to your real users.
I'm currently booking new clients and can typically start within a week. A Production Sprint is scoped in a single 20-minute call, then kicked off immediately.
Production Sprints are a flat project fee scoped to the work. Embedded retainers are monthly and part-time. Advisory is a lighter monthly engagement. The fastest way to a number is a short call describing what you're trying to ship.
Both. I've shipped AI in production at a high-scale platform (ImmuneFi) and built end-to-end agentic products for startups. I work remotely on US hours and also serve Latin American teams in Spanish.
Autonomous agents that take real actions (browser automation, invoicing, sales outreach), document-intelligence and RAG systems with cited answers, and LLM features hardened for production with evaluations, guardrails, and cost control. I work primarily with the Claude Agent SDK, OpenAI, LangChain, and Playwright/MCP.
Let's talk
Got an AI feature that's stuck?
Tell me what you're trying to ship. If I can help, I'll say so on a 20-minute call — and if I'm not the right fit, I'll point you to who is. No pitch deck.
Book a 20-min call →