AI is only one part of the system.

My background is in economics and finance. Research with statistical and econometric methods led me to data analysis, then to machine learning, and from there to the software engineering that makes models useful in production.

Today I work on the layers around the model: APIs, databases, queues, infrastructure, integrations, data pipelines and the architecture that turns an AI capability into a reliable product. The goal is simple: systems that are useful in the real world, not just impressive in a demo.

I like understanding the whole system.

At work that means following a request from the API, through queues, workers and the model, to the database and back out. At home it means a Raspberry Pi, a GPU box, an MQTT broker and a touchscreen face that all have to agree with each other.
Santiago Garcia

Experience

  1. Software Engineer

    HR Agent

    Current

    Production AI platform: backend services, queue-driven workers, LLM pipelines, real-time voice and Azure infrastructure.

  2. AI Developer

    Insightplay

    2025

    Multi-agent systems, FastAPI backends, vector retrieval and cloud deployments on AWS.

  3. Data Analytics Intern

    ProPacífico

    2024

    Interactive dashboards, data validation procedures and statistical analysis of project execution.

Santiago Garcia with the third-place trophy of the first Datathon del Pacífico

Recognition

  • Third place, first Datathon del Pacífico

    Regional data competition organised by ProPacífico. Cali, 2023.

  • Winner, Generative AI hackathon

    A program that generates questions and evaluates reading, speaking and listening proficiency against the CEFR framework.

  • Best Presentation Award

    V Latin American Congress on Social Marketing.

The model is one box.

A request through a typical production AI system. Hover or focus any component to see what it is there for.

cloud, containers, CI/CDeventwebhook, cronAPIcontracts, authqueueback-pressureSQL + vectorssystem of recordworkersscale outLLMvalidationschema checksobservabilitytraces, metricsmessagingSMS, voicecloud, containers, CI/CDeventwebhook, cronAPIcontracts, authqueueback-pressuredataworkersscale outLLMvalidationschema checkstracesmessagingSMS, voice

The model is one box. Everything around it is what turns it into a product.

How I think about engineering

  1. Understand the problem

    Who is it for, what breaks today, and what would "working" look like.

  2. Design the system

    Components, contracts and failure modes before the first line of code.

  3. Build the smallest useful version

    Something real that runs end to end, then add complexity.

  4. Integrate it with the real world

    Existing data, APIs, infrastructure and the people who use it.

  5. Measure, test and improve

    Observe what it actually does, then go around the loop again.

Then back to the first step, with better information.

Where I work in the stack

AI engineering

Models as components with contracts, costs and failure modes.

  • LLMs
  • Generative AI
  • NLP
  • AI agents
  • Local AI
  • Small language models

Software engineering

The code that makes a capability dependable.

  • Python
  • Backend systems
  • APIs
  • Databases
  • System design
  • Software architecture

Cloud & infrastructure

Where it runs, how it ships, how you know it works.

  • Azure
  • Docker
  • CI/CD
  • Cloud infrastructure
  • Production systems

Systems & automation

Connecting software to data, services and hardware.

  • Automation
  • Data processing
  • Integrations
  • MQTT

Education

  • Specialization in Software Engineering

    Universidad Internacional de La Rioja (UNIR)

    In progress

  • Economics and Finance

    Universidad de San Buenaventura, Cali

    2020 – 2024

Certifications

  • Generative AI Engineering with LLMsIBM
  • Advanced Data AnalyticsGoogle

Languages

  • SpanishNative
  • EnglishNative-level proficiency
  • ItalianIntermediate