My CV as a map. Each star is a degree, a job or a project.
How to use the map
Pick a lens to show its stars and lines. Click a star to open its details and its links. The address bar keeps the lens and the star, so you can share the link.
The 4 lenses
EngineeringDegrees and jobs in order
8 stars7 connections
From my BSc at Universidad de los Andes through Bancolombia, Glamper, Imperial and Untap to Tessl. My Formula 1 thesis and the Untap extraction engine branch off the line.
BSc Computing
F1 Prediction
Bancolombia
Glamper
MSc Computing
Untap
Financial Extraction
Tessl
ML systemsProjects with models
6 stars7 connections
My MSc thesis, the 2 LLM systems I built at Untap, EvalLens, LFM and Intus. Each one builds, runs or checks a model.
HPC Regression Detection
Financial Extraction
Investor Intelligence
EvalLens
Large Football Model
Intus
Side projectsBuilt in my own time
5 stars7 connections
This site, the Constellation, EvalLens, LFM and Intus. The site, the Constellation and EvalLens are online. LFM and Intus are in progress.
rendonarango.com
Constellation
EvalLens
Large Football Model
Intus
ResearchDegrees and theses
4 stars4 connections
A BSc at Universidad de los Andes with a thesis on Formula 1 winner prediction, then an MSc at Imperial College London with a thesis on performance regression detection in Nektar++.
BSc Computing
F1 Prediction
MSc Computing
HPC Regression Detection
Numbers
Results from my 2 theses and from my work at Untap.
Years
2018 to 2026
From
Medellín, Colombia
Stars
15
Connections
25
BSc thesis, F1 winner prediction
~93%
About 93% accuracy on race winners, on 70+ seasons of Formula 1 results. The picks made a profit in a backtest against betting odds.
Untap, financial extraction
90%+
Accuracy of the fine-tuned LLM pipeline that pulls KPIs, targets and cashflows from financial documents.
Untap, ingestion time
−80%
Cut in the time clients took to load their data, against the old manual process.
MSc thesis, HPC triage
Days to under 2 hours
Median time to triage a performance regression in Nektar++. A variational autoencoder over performance counters runs in GitLab CI on each commit.