First engineering role
BinaryTools (Bogotá)
Backend Java for Univex — a secure platform used by Colombia's Counterintelligence School and National Army.
Camilo Pestana · PhD
Head of Data Science & ML at a $4.15B gaming company.
Built production fraud detection worth $230M+ in year one. Contributed to two FDA-cleared cardiac AI models. DARPA-funded PhD, 5 IEEE papers, and a side-stack of open-source agentic tooling and multiplayer games — written by the same person who owns the strategy.
Selected work
Agentic tooling, AI security, medical imaging, multiplayer games, full-stack clones, research code — all shipped. Hover any card to see it in motion.
Agentic + LLM + CV fraud system at $4.15B scale.
Two FDA 510(k)-cleared models for coronary artery disease.
Live observability for Claude Code agents.
PhD-grade red-team curriculum — open source.
Self-guided AI fundamentals → modern GenAI.
AI agents that deliver real work, on demand.
Realtime AI whiteboard. Generated from one prompt.
Full-stack social app shipped by Claude from one line.
Brotato-style arena co-op. AI-generated art + SFX.
Balatro-inspired multiplayer dice deckbuilder.
Slay the Spire–inspired slot roguelike.
Computer vision for X-ray diagnosis. 314★ on GitHub.
Premium modular spaces — marketing site.
Rust CLI for multi-agent orchestration.
The path
Fifteen years across software engineering, ML research, medical AI, fraud detection, and adversarial security. Scroll horizontally — the pivots that mattered.
BinaryTools (Bogotá)
Backend Java for Univex — a secure platform used by Colombia's Counterintelligence School and National Army.
Colombia · USA · Australia
Engineering for Southwest Airlines (Netflix OSS microservices), National Geographic, Disney, ESPN, Royal Caribbean, Bank of Bogotá.
Years before AI went mainstream. Bet the career on it and never looked back.
Real-time roadside hazard detection with deep CNNs. Won the WA Ministry of Data Hackathon (40+ teams) and $21K SPURonWA innovation grant.
University of Western Australia
DARPA-funded adversarial ML research. 5 IEEE papers (CVPR, WACV, 3DV, IJCNN). Cloud Computing lecturer.
Led $800K AUD Colombian Government project for X-ray / CT anomaly detection. Featured in Colombian national media.
UWA · Australian Government
ACS 1962 Medal Finalist (top 5 PhD thesis in WA). Australian Global Talent Visa for outstanding expertise in AI.
Senior Data Scientist
Contributed to two FDA 510(k)-cleared AI models for coronary artery disease (stenosis estimation + plaque classification).
$4.15B revenue · Perth
Built and shipped a production fraud detection system combining Agentic AI, LLMs, and CV — $230M+ USD ROI in year one.
Operation Blackout — 41st globally out of 796 teams. Solved Neural Detonator (the hardest AI/ML challenge) by reverse-engineering a hidden payload in a Keras model.
VGW
Promoted to unify AI/ML Engineering and Data Science into one ~20-person enterprise function. Owning strategy, governance, and delivery across all AI/ML investment.
VGW
Modernising the platform from Databricks to AWS (SageMaker, Bedrock, Iceberg, MSK). Productionising agentic AI on Bedrock with Guardrails + LangGraph. Leading a ~20-person enterprise AI/ML + Data Science function.
Domains
Gaming, defense, medtech, academia, AI security, oil & gas — shipping in each with the same hands.
$230M+ ROI fraud detection. Agentic AI, LLMs, CV at $4.15B scale.
DARPA-funded adversarial ML research. Secure systems for military intel.
Two FDA 510(k)-cleared AI models. 3D CNN cardiac imaging. Biosensor ML.
PhD in AI (UWA). 5 IEEE publications. Cloud Computing lecturer.
HTB AI Red Teamer. 2nd in Australia CTF. Open-source security course.
First open-source DS/ML curriculum for Oil & Gas industry.
Stack
What I reach for, day to day — battle-tested across regulated clinical, fintech, defense, and gaming workloads.
Productionising LangGraph + Bedrock Guardrails at VGW.
5 IEEE papers on adversarial ML; PyTorch since 2018.
Two FDA-cleared cardiac imaging models at Artrya.
DARPA-funded PhD. HTB CTF 2nd in AU, 41st globally.
Modernising Databricks → AWS at $4.15B scale.
Tarsa, Inkwell Dice, this site — all Claude-Code-driven.
Research & recognition
Five IEEE papers from the PhD years on adversarial ML, plus a shelf of recognitions across academia, industry, and security competitions.
Exceptional Impact Award, VGW
Revenue growth and shaping VGW's AI & Data strategy.
HTB CTF: 2nd in Australia / 41st Globally
796 teams. Solved Neural Detonator (hardest AI/ML challenge).
Australian Global Talent Visa
Granted for outstanding expertise in AI and Data Science.
ACS 1962 Medal Finalist
Top 5 PhD thesis in Western Australia.
Distinguished Alumni of the Year
Catholic University of Colombia.
COVID-19 AI Detection Project
$800K AUD Colombian Government project. National media.
AI Red Teamer Path
Hack The Box Academy · 2026
Co-developed with Google's Secure AI Framework. <43% completion rate.
Global Cyber Skills CTF: Operation Blackout
Hack The Box · 2025
2nd in Australia · 41st globally out of 796 teams.
Agentic AI Specialisation (7 courses)
DeepLearning.AI · 2025
Eval & Govern Agents · DSPy · AutoGen · LangGraph · LangChain.
RSNA Medical Imaging AI (Advanced)
RSNA · 2023
RSNA Medical Imaging AI (Foundational)
RSNA · 2023
ML in Production
DeepLearning.AI · 2022
AI for Medical Diagnosis
DeepLearning.AI · 2020
About
I'm Camilo. Perth, Australia. Started writing Java for Colombia's National Army in 2011, pivoted to ML in 2017 before it was cool, finished a PhD on adversarial computer vision in 2022, and now run the central AI/ML & Data Science function (~20 people) at VGW — a $4.15B gaming company.
Day job is C-suite advisory, AI governance, and shipping production systems at scale (fraud detection, agentic AI on Bedrock, modernising Databricks → AWS). Off-hours, I ship things: agent observability tooling, multiplayer games on Godot, full-stack apps built end-to-end with Claude Code, an open-source AI security course used by red-teamers.
Same person designs the strategy and writes the code.
That's the whole pitch.
Say hi
Roles, collaborations, advisory, interesting research, weird ideas — all welcome.