Building production AI systems

Alejandro Alvarez.AI Engineer & Full Stack Developer

I build production LLM systems — multi-agent architectures, MCP servers, and serverless AI pipelines on AWS — with the full-stack depth to take them from prototype to product.

View projects

Get in touch

Medellín, Colombia · UTC−5 · English C1

100K+
calls scored by LLM pipeline: −35%
LLM error rate via prompt engineering: 500+
business locations automated: 4+
years shipping AI to production

01 — About

From prototype to product

Alejandro Alvarez Uribe · Medellín, Colombia

I'm an AI Engineer who ships LLM systems that hold up in production — multi-agent architectures, MCP servers, and serverless AI pipelines on AWS (Bedrock/Claude, Step Functions, Lambda). My full-stack foundation in NestJS, React, and PostgreSQL means I don't just prompt models: I design the architecture, integrate it end to end, and measure it.

Production AI, not demos

Real LLM pipelines on AWS serverless with automated scoring over 100K+ calls — not notebooks.

Multi-agent & MCP servers

Hands-on experience architecting agent systems and MCP servers for real business workflows.

Prompt engineering with metrics

−35% error rate on agent-driven operations. Measurable optimization, not anecdotes.

End-to-end ownership

Architecture, backend, frontend, infra, and CI — one profile covers idea to production.

02 — Experience

Where I've shipped

Four years taking generative AI from idea to deployed systems for U.S. and LATAM clients.

Fullstack Developer · Olelo Intelligence

Nov 2025 — Presentcurrent

LLM-powered call intelligence for customer service at automotive centers.

Built the TMA/IMA LLM analysis pipeline on AWS serverless (CDK, Step Functions, Lambda, Bedrock/Claude), scoring 100K+ customer-service calls in production.
Developed multi-tenant NestJS + PostgreSQL backends and React dashboards across five core repositories.
Secured the platform with AWS Cognito JWT auth, Redis caching, and S3 + CloudFront media delivery.
Designed SMS-invoice-to-call transaction matching to defend paid-for-service KPIs, leading cross-functional product decisions.

AWS CDKBedrock / ClaudeStep FunctionsNestJSPostgreSQLReact

AI Engineer · Insighty AI

Apr 2025 — Mar 2026

Multi-agent systems and MCP servers powering AI-driven business workflows.

Architected multi-agent systems and MCP servers for scalable automation and real-time decision support.
Reduced LLM error rates by 35% through systematic prompt engineering, improving task completion speed for agent-driven operations.
Designed end-to-end AI architectures integrating NestJS services, PostgreSQL storage, and orchestration layers for high-volume, low-latency workloads.
Aligned cross-functional teams on AI infrastructure strategy, accelerating delivery of deployed AI solutions.

Multi-agentMCP serversPrompt engineeringNestJSPostgreSQL

Fullstack AI Developer · Geeks5G

Mar 2025 — Sep 2025

Generative-AI content automation for local businesses across the U.S.

Automated content creation for 500+ Google My Business locations with generative AI, increasing posting frequency by 70%.
Built fullstack services with NestJS and PostgreSQL, orchestrating workflows for scraping, keyword scoring, and scheduled publication.
Developed AI tools for localized posts, review responses, and summaries, integrating OpenAI models with JSONB-backed data flows.
Led backend infrastructure for a scalable content-automation system serving multiple U.S.-based clients.

OpenAINestJSPostgreSQLn8nCI/CD

Fullstack AI Developer · Capital Pocket

Oct 2023 — Mar 2025

AI automation and microservices for fintech and public-sector clients.

Developed AI solutions that automated 35% of manual tasks and cut processing time by 50% across two companies.
Built generative-AI systems and conversational agents for multiple organizations as a backend developer (TypeScript).
Integrated APIs, managed databases, and built microservices with a focus on scalability and performance.
Led a development project for a Colombian public entity, improving efficiency across marketing, administrative, and operations teams.

TypeScriptNode.jsMicroservicesLangChainMongoDB

03 — Projects

Selected work

AI agents, automation platforms, and products in production — a sample of what I've built and shipped.

Pocki — WhatsApp AI Agent

Conversational AI agent living on WhatsApp Business: information retrieval, reminders, content generation, and image processing, with MongoDB-backed context for personalized interactions.

LangChainWhatsApp APINode.jsMongoDB

Google Rank AI

AI platform that manages and optimizes Google My Business listings at scale — automated content, keyword scoring, and data-driven decisions to improve local search rankings.

Generative AINestJSPostgreSQLSEO automation

Cyberix

AI agent platform for customer interactions and business automation: cross-platform integrations, real-time analytics, and a documented API for scalable deployments.

AI agentsREST APIAnalytics

Metroparques Admin Portal

Administrative platform for a Colombian public entity — centralizes tasks and streamlines workflows for admin, marketing, and management teams. Led delivery with Scrum.

ReactNode.jsScrum