Skip to content
← Back to the worldAbout

Camilo Pestana

AI/ML leader · researcher · engineer

Perth, Western Australia

Hands-on AI/ML and data science leader with a PhD in AI (adversarial machine learning and computer vision) from the University of Western Australia. He builds production AI — from fraud detection and governed agentic GenAI to clinical imaging models — and leads the teams that run it.

Experience

  1. Principal Delivery Consultant – AI/ML

    Upcoming

    Amazon Web Services (AWS) ·

    Joining AWS — November 2026

    Camilo's next chapter: he is joining Amazon Web Services Professional Services as a Principal Delivery Consultant in AI/ML, with a start date of 16 November 2026.

  2. Head of Data Science & ML

    Current

    Virtual Gaming Worlds (VGW) · Perth, Australia ·

    Promoted to unify AI/ML Engineering and Data Science at VGW into a single enterprise function, owning strategy, governance and delivery across the company’s AI/ML investment.

  3. Data Science Manager

    Completed

    Virtual Gaming Worlds (VGW) · Perth, Australia ·

    Built and led VGW’s Data Science & Analytics team, staying hands-on as architect of core systems, and was promoted to Head of Data Science.

  4. Senior Data Scientist

    Completed

    Artrya · Perth, Australia ·

    Led machine-learning model development and MLOps for clinical AI that helps doctors find coronary artery disease on heart CT scans, working directly with cardiologists and radiologists.

  5. Head of Data and Software (joined as Senior Data Scientist)

    Completed

    VitalTrace · Perth, Australia ·

    Joined VitalTrace as Senior Data Scientist (tech lead) and was promoted to Head of Data and Software, building the data, software and ML capability from zero for the company’s first-of-its-kind continuous fetal lactate monitor.

  6. Lecturer & Researcher (AI / Computer Vision)

    Completed

    The University of Western Australia (UWA) · Perth, Australia ·

    At the University of Western Australia Camilo lectured Cloud Computing (AWS) and researched adversarial machine learning for computer vision in a DARPA-funded program.

  7. Software Developer & ML Engineer

    Completed

    HIVO · Perth, Australia ·

    Built ML and computer-vision features — facial recognition, image search and asset analytics — into a cloud digital-asset-management platform.

  8. Co-founder & Chief Data Scientist

    Completed

    SmartVision AI · Perth, Australia ·

    Co-founded a computer-vision startup building real-time roadside hazard detection from video, in partnership with Main Roads Western Australia.

  9. Software Engineer

    Completed

    Globant · Colombia, USA, Australia ·

    Software consultant delivering production features for enterprise clients across Colombia, the United States and Australia.

  10. Backend Java Developer

    Completed

    BinaryTools · Bogotá, Colombia ·

    First professional engineering role: backend features for a secure information-management system used by government and defence clients in Colombia.

Education

  • PhD in Computer Science (AI)

    The University of Western Australia · Perth, Australia · 2019 – 2022

    Camilo’s PhD at UWA studied non-traditional adversarial machine learning for computer vision: attacks, defences and 3D adversarial textures.

  • Honours in Computer Science, First Class

    Edith Cowan University · Perth, Australia · 2017

    First-class honours with high distinction; thesis modelled influenza-like illness across Australian regions from geotagged Twitter data.

  • BEng Software & Systems Engineering

    Catholic University of Colombia · Bogotá, Colombia · 2007 – 2013

    Undergraduate thesis designed a web information system linking environmental, social and economic criteria to improve social-housing prototypes in Bogotá.

Skills

Agentic & multi-agent GenAI
LLMs · Multi-agent system design · LangGraph / AutoGen / DSPy · Tool calling · RAG & vector databases · Prompt engineering · LLM fine-tuning & evaluation · AWS Bedrock + Guardrails · Agent evaluation & safety
Evaluation, interpretability & research
Model interpretability · Evaluation sets & benchmarks · LLM & agent evaluation · Robustness & failure-mode analysis · Adversarial ML · AI safety & red-teaming · Experimental design
Machine learning & languages
Deep learning · Python · PyTorch · TensorFlow · scikit-learn · LightGBM · Active / semi-supervised learning · SQL · R
Computer vision
CNNs / 3D CNNs / VLMs · Object detection / OCR · Segmentation · Medical imaging
MLOps & cloud (AWS)
Bedrock · SageMaker · Glue · Apache Iceberg · Kafka (MSK) · MLflow · Weights & Biases · Airflow · Docker · Snowflake · Spark / Databricks · Grafana
Enterprise architecture & software
Enterprise AI & cloud transformation · Cloud-native architecture (AWS) · Solution architecture · Enterprise GenAI adoption patterns · Microservices & REST APIs · CI/CD · Java / TypeScript / Node.js
Leadership
Team leadership & mentoring · Stakeholder management · C-suite & board communication · Strategic thinking · Cross-functional collaboration · Trusted technical advisor · Conference talks

Publications

  1. Assistive Signals for Deep Neural Network Classifiers · IEEE/CVF CVPR Workshops, 2021 · code
  2. Attacking Image Classifiers To Generate 3D Textures · IEEE 3DV, 2022 · code
  3. Transferable 3D Adversarial Textures using End-to-end Optimization · IEEE/CVF WACV, 2022 · code
  4. Defense-friendly Images in Adversarial Attacks: Dataset and Metrics for Perturbation Difficulty · IEEE/CVF WACV, 2021 · code
  5. Adversarial Attacks and Defense on Deep Learning Classification Models using YCbCr Color Images · IEEE IJCNN, 2021 · code

Selected projects

  • NOVA: catching fake mail-in entries

    Virtual Gaming Worlds (VGW) · 2024 – 2026

    A computer-vision system Camilo proposed and led at VGW that caught fake sweepstakes mail-in letters, keeping the free entry route fair at a volume manual checks could not handle.

    As of Aug 2025 NOVA was checking ~3M entries a week, with ~27% of entries flagged as potential fraud each week and ~$4.05M saved per week. Over its life, from 2024 into 2026, it saved millions of dollars and replaced manual checks that could not keep up.

  • Salix: AI that reads heart scans

    Artrya · Nov 2022 – Apr 2024

    At Artrya, Camilo led the machine-learning work behind models in the Salix family of products, which help doctors find and grade narrowed heart arteries on a CT scan.

    The models feed Artrya’s Salix products, and the Salix product family includes FDA 510(k)-cleared software. Those clearances belong to Artrya and its products; Camilo’s part was the model development, clinical validation and submission work on stenosis estimation and plaque classification.

  • VitalTrace: a continuous fetal lactate monitor

    VitalTrace · May 2021 – Nov 2022

    At VitalTrace, a MedTech startup making a first-of-its-kind continuous fetal lactate monitor, Camilo built the data, software and machine-learning capability from zero.

    The company gained its own data platform and analysis software in place of external vendors. The continuous fetal lactate monitor the models supported later secured FDA De Novo designation, a company milestone.

  • Structured AI for vision

    Independent (github.com/elcronos/agentic) · Ongoing

    A method Camilo proposed for building image-inspection AI quickly: program a vision-language model with DSPy instead of hand-writing its prompt, and let it improve from a small set of labelled examples.

    In his public proof of concept on screw images, each round of optimisation and active learning raised the score on a small held-out test set, until it got all but one test image right, starting from only a handful of labelled examples per class.

  • AI Evals for Engineers

    Independent (course author) · Ongoing

    A free, practical course Camilo wrote on testing AI apps and agents, so teams can tell whether their AI really works before their customers find out.

    The course is public and free, and it gives teams a repeatable way to decide when an AI feature is good enough to ship.

  • Games, and this island, built with AI agents

    Independent (capcarde.itch.io) · Ongoing

    Camilo builds browser games, and built this island, by directing AI coding agents and automated loops, with AI-generated art and sound.

    Three published browser games and this playable portfolio, each built end to end with AI agents under his direction.

  • SmartVision AI: spotting road hazards

    SmartVision AI · Oct 2017 – Jun 2019

    Camilo co-founded SmartVision AI, a startup that used AI to spot roadside hazards in video in real time, in partnership with Main Roads Western Australia.

    The startup won the WA Ministry of Data Hackathon and received a SPURonWA government innovation grant.

  • Free courses and tools

    Independent (github.com/elcronos) · Ongoing

    Free courses and tools Camilo publishes in the open: an AI Security course, a self-guided AI course, Tarsa for watching AI agents at work, and CraftBench for testing AI on real engineering tasks.

    All of it is public and free to use.

Honours & community

  • ACS 1962 Medal finalist (PhD thesis) · Sep 2022His PhD thesis was a finalist for the ACS 1962 Medal, recognised among the top PhD theses in Western Australia.
  • Australian Global Talent visa · Sep 2022Awarded residency under Australia’s Global Talent program for expertise in AI and data science.
  • Exceptional Impact Award, VGW · Aug 2025VGW recognised his exceptional impact on the business and his part in shaping the company’s AI and data strategy.
  • Hack The Box Operation Blackout CTF (AI/ML security) · May 2025Competed in Hack The Box’s Global Cyber Skills CTF focused on AI/ML security, solving its most complex AI/ML challenge by reverse-engineering a hidden payload in a Keras model.
  • COVID-19 AI detection project · Jun 2020Led an AI project detecting anomalies in chest X-ray and CT images, funded by the Colombian Government and featured in national media.
  • Organiser & speaker, Perth Machine Learning Group · Feb 2019 – presentOrganises and speaks at the Perth Machine Learning Group, including the talk “Breaking AI Systems” on AI security and adversarial attacks.

Elsewhere