The person behind the systems
About Camilo
Applied AI Engineer. Immigrant. Runner. I build things that work and I do not stop until they ship.
The Real Version
I studied petroleum engineering in Colombia. Good at it, but knew I wanted something different. Moved to New York with a plan: get a master's degree, learn to build software, figure out the rest later.
“Figure out the rest later” turned out to mean bartending four nights a week at a cocktail bar in Manhattan while taking a full course load at Baruch. The other hours went to building things. Not tutorials. Not demos. Actual applications that real people use.
The project that changed everything was Invoz. I wanted to build a speech scoring system, so I read 46 research papers on audio processing and taught myself signal processing from scratch. That project became an 11-dimension pronunciation scorer running Whisper, wav2vec2, Parselmouth, and Silero VAD in production. It taught me that going genuinely deep on a hard problem is more valuable than being broadly familiar with easy ones.
Then I built Holus -- a 32-agent autonomous system with Redis pub/sub, silo isolation, guardrails, and self-improvement loops. That project confirmed what I suspected: the hard problems in AI are not the models. They are the systems around them. Orchestration, reliability, observability, failure recovery -- the stuff that does not make good Twitter threads but determines whether your system works at 3 AM.
I have shipped 10+ production systems and counting. I run 5 days a week, I am trying to hit every street in Astoria, and I am looking for a team where the bar is high and the problems are real.
Timeline
Petroleum Engineering, Colombia
Graduated. Learned to think in systems, constraints, and optimization under pressure.
NYC -- MS Business Analytics, Baruch
Moved to New York with no safety net. Bartended four nights a week. Built ML projects every other waking hour.
Invoz.ai -- From Papers to Production
Read 46 research papers. Built a speech ML pipeline that scores pronunciation across 11 dimensions. First time I felt like a real engineer.
Holus -- The Systems Problem
32 autonomous agents. Redis pub/sub. Silo isolation. Proved to myself that the hard part of AI is not the model -- it is everything around it.
10+ Production Systems, Still Hungry
Shipping audio ML, agent infrastructure, and full-stack applications from Queens. Looking for a team that builds things that matter.
What I Build With
Audio/Speech ML
Whisper, wav2vec2, Parselmouth, Silero VAD. I built a production pronunciation scoring system -- not a wrapper around an API.
Multi-Agent Systems
LangGraph orchestration, Redis event bus, health preflight, observability, silo isolation. I think about agent reliability the way SREs think about uptime.
Full-Stack & Data
TypeScript, Next.js, FastAPI, PostgreSQL, Redis. I build the entire stack because waiting for someone else to unblock you is a luxury early-stage does not have.
Business-Aware Engineering
MS in Business Analytics means I think about cost-per-inference and unit economics at the architecture stage. Not after launch.
How I Work
Systems over motivation
I do not rely on feeling motivated. I build structures that produce output regardless -- automated pipelines, daily commits, public accountability.
Depth over breadth
46 papers for one feature. 32 agents for one system. I would rather master one hard problem than skim ten easy ones.
Ship then measure
Working software is the only credible argument. I ship first, measure impact, then iterate based on what actually happened -- not what I imagined would happen.
Honest about gaps
I do not perform expertise I do not have. If I do not know something, I say so, then I go learn it. That is faster than pretending.
Ask Me Anything
An AI that knows my background. Try it.
Hey! Ask me anything about Camilo -- his skills, his story, what he built, or what he is looking for.
Looking for Applied AI Engineer roles. Available now. Based in NYC.