Most recently
Founding engineer at Enttor
Studied at
Universidad de los Andes

Hello

I'm a founding engineer in Bogotá, looking for the next thing. I ran engineering at Enttor, where I built the paid ads platform on Meta and TikTok and the Slack agent that runs it. Before that I worked on an e-commerce marketplace at Samsam. I have a master's from Universidad de los Andes.

Bogotá, Colombia13:07

Experience

The jobs, and the school around them.

Product engineer at Samsam, then founding engineer at Enttor. I did the master's while working the first job.

Click a row to open it

It started as the outbound engine, browser agents sourcing prospects and drafting the DMs. It ended somewhere else: Meta and TikTok campaign data normalized onto one metric schema, a rules engine that returns scale, pause or archive with a reason, and a Slack agent media buyers drive in plain language across 165 tools. The last product generated the ads outright, then took them back apart into editable layers on GPU workers.

Products shipped
Three
Ad generation
1 hr to under 5 min
Agent tools, Meta and TikTok
165
Team
Six
Built with
TypeScript, Next.js, Supabase, Playwright, Claude, Modal

Backed byNeo

A React Native and Expo app for shoppers and a Next.js dashboard for merchants, on NestJS services with Prisma and Postgres. I shipped both sides.

Shipped
Both sides of the marketplace
Platforms
iOS, Android, web
API response times
10% faster
Production alerts
Routed by severity
Built with
React Native, Expo, Next.js, NestJS, Prisma, Postgres

Deploys were manual when I arrived. I moved the pipeline to Azure and built the CI/CD that released to every client server on its own. I owned a Python and Flask reporting service running scheduled queries across client databases, and built the Power BI dashboards clients used to track headcount, turnover and payroll cost, replacing reports that had been assembled by hand.

Deploys
Manual to automated
Release targets
Every client server
Reporting service
Scheduled queries
Dashboards
Headcount, turnover, payroll
Built with
Python, Flask, Java, Spring Boot, Azure, Power BI

It starts with data that does not fit a table, semi-structured, geolocated, high-velocity, multimedia, and asks what it is worth to whoever is holding it. Most of the time goes on governance, security and data quality, because that is what any decision made on top of it rests on. Coursework ran from applied data science and machine learning through recommender systems, knowledge engineering and privacy.

Length
Four semesters
Studied
Part-time, while working
Department
Systems and Computing
Field
Data at scale
Coursework
Applied data science, Big data, Machine learning, Recommender systems, Knowledge engineering, Security and privacy

Mathematics for the formal side, management for the business side. Both ran alongside the undergraduate degree and closed before graduation.

Mathematics
For the formal side
Management
For the business side
Taken
Alongside the B.Sc.
Closed
Dec 2022, a year early
Areas
Mathematics, Management

Systems, algorithms and applied ML. Side projects in Python, TypeScript, Java and Swift along the way.

Duration
5 years
Thesis
Crops from satellite imagery
Accuracy
98.4%
F1
0.991

Honors & certifications

A national distinction and two AWS certifications.

  • 2018

    Andrés Bello National Distinction

    Ministry of Education, Colombia

    Awarded by the Ministry of Education for a top-ranked score on Colombia’s national high-school exit exam (Saber 11).

  • 2022

    AWS Academy Cloud Developing

    Certification, Amazon Web Services

    Building, deploying, and scaling apps on AWS: IAM, Lambda, DynamoDB, S3.

  • 2021

    AWS Academy Cloud Foundations

    Certification, Amazon Web Services

    AWS fundamentals: core services, security, architecture, and pricing.

Selected work

Things I've built.

Product, infrastructure, and a few things on the side.

AI Ad Studio, ads that build themselves.

Enttor's last product. Give it a brand's website and it scrapes the logos, colors and fonts, writes ad concepts with Claude, renders the images, then checks its own work with a vision pass. The harder half runs backwards: OCR and segmentation take a finished ad apart into editable layers, matching every line of text against 311 font families, so a media buyer edits on a canvas instead of rebuilding from scratch. Both vision workers run on GPUs.

Next.js, Python, PyTorch, Modal, Supabase

Slack Agent

A Slack bot media buyers talk to like a coworker. 165 tools across the Meta and TikTok ad APIs. Accuracy falls off long before 165 tools fit in context, so a router narrows every message to a shortlist before Claude sees it.

Claude tool use, Meta and TikTok APIs, pgvector

Enttor Platform

Two years in one repo. It sourced prospects from Instagram and LinkedIn, wrote personalized DMs and tracked every reply in one inbox, then in 2026 was refocused on paid ads: Meta and TikTok insights on one metric schema, scored against per-client thresholds. Row level security isolates every company from the rest, so a handler that forgets its filter returns nothing.

Next.js, Supabase, Postgres RLS

MakeMotionGraphics

Text to looping motion graphics. Claude writes a self-contained HTML scene, screenshots it in headless Chrome at three timestamps to read back its own output, then patches what is wrong. Only a scene that clears the composition gate gets recorded to MP4.

Agentic loop, Puppeteer, ffmpeg

LinkedIn Editor

A chat-first editor for LinkedIn posts. You paste a draft and argue with Claude about the hook until it lands, grounded by cosine search over a corpus of posts that actually performed. Three variations stream in parallel, one per hook angle.

pgvector RAG, Next.js

Browser Automation Infrastructure

The backend behind the outreach: Node services on a Mac Mini that keep authenticated browser sessions alive on four platforms, with Redis queues pacing outreach in minutes rather than milliseconds. Two teammates led it, I contributed the Instagram pipeline and session handling.

Playwright, Redis, Postgres

Samsam

Before Enttor. A shopping marketplace: a React Native and Expo app for shoppers and a Next.js dashboard for merchants, on NestJS services with Prisma and Postgres. I shipped both sides.

React Native, Expo, NestJS, Prisma
The deck

The parts a CV leaves out.

Nine photographs from the last few years.

Moments

Pick a card

Click or Esc to let it go

Console

Try the shell. Or ask the AI a question.

Type a command: help, whoami, experience… Want to ask something else? kevin "your question" sends it to the AI.

kevin@gamez, ~ready
welcome, a small shell for this site. type help for commands, or ask the AI directly: kevin "what does enttor do?"
kevin@gamez~$

Let's find out if we should work together.

A 30-minute call, no deck. Bring the messy version of the problem. Or just say hi.

Open to founding-engineer roles