Cerberus Public terminal
José Ignacio
Concha Araya
AI Engineer & Tech Lead Santiago, Chile
I build AI systems that survive contact with production, not demos: a five-model ensemble that prices copper in under 15 ms, a Rust trading engine that retrains on GPU every ten seconds, and multi-agent compliance pipelines that replace weeks of manual regulatory review for Chilean banks. Physicist by training, systems engineer by trade — I lead the team, design the architecture, and write the production code. Almost all of it lives in private repositories, which is exactly the problem this page exists to solve: the code stays closed, the volume does not.
- Portfoliojose.cerberus.cl
- Companycerberus.cl
- LinkedInjose-ignacio-concha-araya
- GitHubgithub.com/l0rdbarcsacs
Five sections. The geometry behind them is drawn from the same JSON the text is, and every figure is regenerated daily. Nothing on this page is decorative.
Contribution city
Contribution city
The twelve-month contribution calendar, extruded. One box per day on the same 53 × 7 grid GitHub uses — a column is a week, a row is a weekday. Height is log1p(count) / log1p(p98): the 98th percentile by nearest rank rather than the maximum, because one very heavy day would otherwise press the rest of the year flat into the floor. Colour is driven by that same normalised height instead of GitHub’s quartile bucket, so the two can never disagree with each other.
Loading the contribution figures.
| Month | Contributions | Active days | Peak day |
|---|
Language distribution
Languages
Loading the language figures.
The bars are log-scaled, and they have to be: one language holds well over ninety percent of the bytes, so a linear bar is a single full-width block and eleven invisible ones. The exact linear share is in the table below the bars, where a number belongs. Repository count is printed next to each language on purpose — it measures breadth, which is usually what you are actually trying to infer, and bytes measure only how much was typed.
| Language | Bytes | Share | Repositories |
|---|
Systems in orbit
Systems
Five things worth explaining. Every figure below is the same one published in the CV and on the profile README — none of it was invented to make the visualisation look busier. Select one to bring it to the front of the scene.
The rig & the route
Rig
Everything above is built, trained and benchmarked on hardware I own and tune myself. The GPU numbers on this page are measured on the first row of this table, not quoted from a datasheet.
| Node | Specification | Role |
|---|