Felipe.Florez
CV ↓

Founding Engineer at Emilia · AWS Solutions Architect · Medellín, CO

I build agent & LLM systems that run in production.

At Emilia, I build AI products end to end—from architecture and integrations to automation and infrastructure—with speed, scalability, reliability, and user experience in mind.

Portrait of Felipe Florez
57→80%
First-version resolution
~40%
Cloud cost reduction
6+
Years shipping software
AWS certifications

About

I turn LLM prototypes into systems teams can rely on.

I'm a Founding Engineer at Emilia in Medellín, helping build the product and technical foundation from the ground up. I work across product engineering, architecture, APIs, integrations, automation, and infrastructure to take reliable AI systems into production.

I work closely with the founders on product strategy, technical priorities, and execution. My electronic engineering and IoT background taught me to design for real-world constraints; today I pair that reliability mindset with Spec-Driven Development and AI-assisted workflows to move fast without making the system opaque.

Capabilities

What I work with

LLM Systems (Production)

  • RAG
  • Agent Orchestration (LangGraph)
  • Amazon Bedrock
  • Pinecone
  • LiteLLM (model routing)
  • Tool Execution / MCP
  • Guardrails
  • Evaluation & Observability

AI-Assisted Development & SDD

  • Claude Code
  • Cursor
  • Devin
  • Spec-Driven Development
  • Spec Kit
  • OpenSpec
  • Anthropic Claude API
  • OpenAI API
  • Prompt Engineering
  • Agent Scaffolding

Cloud & Backend

  • AWS Lambda
  • DynamoDB
  • SQS
  • API Gateway
  • Bedrock
  • Supabase
  • Vercel
  • Docker
  • CI/CD
  • Event-Driven Systems
  • Microservices
  • Serverless

Engineering Foundations

  • Python
  • REST APIs
  • System Integration
  • Data Modeling
  • PostgreSQL
  • JWT/Auth

Experience

Where I've shipped

  1. Aug 2026 – Present

    Remote · Medellín, Colombia

    Founding Engineer

    Emilia

    • Build the product and technical foundation for real-time restaurant AI, spanning end-to-end features, architecture, APIs, integrations, automation, and infrastructure.
    • Partner with the founders on product strategy, technical priorities, and execution, balancing speed with scalability, reliability, and user experience.
  2. Feb 2026 – Aug 2026

    Remote

    R&D Engineer, AI

    SoftServe

    • Design agent-based SDLC workflows (Devin, Claude Code, Cursor) for enterprise clients in finance and retail.
    • Lead AI-adoption workshops and prototype agentic patterns and evaluation frameworks for emerging LLM use cases.
  3. Feb 2025 – Feb 2026

    Remote

    AI Engineering Lead

    NOUSWARE

    • Lifted first-version resolution from 57% to 80% and cut cloud cost ~40% with tiered model routing.
    • Built a serverless agent orchestrator on AWS Lambda + Bedrock with RAG (Pinecone), DynamoDB-backed state, and secure, typed tool execution (Pydantic).
  4. Sep 2023 – Feb 2025

    Remote

    Technical Consultant (Integration Engineer)

    Ivanti

    • Delivered 18+ production automation workflows and integrations across 7+ external systems.
    • Shipped GenAI-based ticket triage that cut manual processing time ~35%.
  5. Feb 2023 – Sep 2023

    Remote

    Support Engineer (L2)

    Ivanti

    • Resolved 80+ complex production incidents, improving MTTR ~28%.
    • Authored runbooks and known-issue guides that reduced ticket escalation ~30%.
  6. Jan 2020 – Feb 2023

    Remote

    Founder & IoT Solutions Architect

    Floes Connecting

    • Built and deployed an end-to-end IoT solution (MQTT/EMQX on AWS, C++ firmware on ESP-01) running in production.
    • Owned everything from circuit design and a PHP + SQL platform to on-site deployment and user training.

Projects

Weekend builds

Every weekend I ship a small project with Spec-Driven Development: a live demo on its own subdomain, a public repo, and a short case study.

Guardrails
▶ PlayableLive

Canary

A browser game where you are the attacker: sneak a prompt injection past a guardrail that runs 100% on your device. A layered pipeline (normalization, heuristics, a 22M ONNX classifier) returns BLOCKED or PASS with a confidence score and the layer that caught you.

  • Transformers.js
  • ONNX
  • Llama Prompt Guard 2 (22M)
  • React
  • TypeScript
Agents
ToolLive

Mi Hogar

A single-user finance tracker you run by chatting with a Telegram bot. An OpenAI agent with typed tools turns text, voice notes and receipt photos into transactions; a Next.js dashboard covers reports, budgets, debts and Colombian payroll.

  • Next.js 16
  • TypeScript
  • Supabase
  • OpenAI
  • Telegram Bot API
Algorithms
▶ PlayableLive

Indetectable

A daily browser game: forge one stroke that blends into the machine's. An algorithmic detective tries to catch the human intruder. You win if it can't.

  • React
  • TypeScript
  • Vite
  • Canvas 2D
  • perfect-freehand

Certifications

Credentials

Education

B.S. in Electronic Engineering

Universidad de Antioquia · Medellín

2017 – 2024