George Ntakakis

George Ntakakis

I'm George Ntakakis, an engineer and technology leader with more than 14 years across software, cloud and AI systems in healthtech, pharma and regulated research. I build agentic AI and machine learning systems for healthcare, and the teams that ship them. I was employee number one and later CTO of Augmented Training Systems in Austin, where we went from zero to a revenue business on programs won from the U.S. Department of Defense and the State Department. I led 16 people across 3 teams, hired 22, and we shipped around 20 simulation training products for hospitals and U.S. embassies worldwide. Before that I was at Microsoft Innovation Center Greece, running the cloud data lifecycle of a European Horizon 2020 Parkinson's study, and earlier I co-founded MicroView, a medical-imaging R&D venture that won two national awards and reached the Microsoft Imagine Cup global semifinals.

Today I work hands-on on agentic AI for a global pharmaceutical client through NewPage Solutions, and this summer I designed the agentic operations layer of a proptech company, rehearsed on 14,000 real messages before launch. I am finishing a PhD at the Aristotle University School of Medicine on VR, biosensors and AI for surgical training, with the degree due in the first quarter of 2027, on a national fellowship from the Hellenic Foundation for Research and Innovation, after an MSc in Medical Informatics at the same university, with 21 publications so far. The graph on the right is my own set of agents, from a rental property to my research and my health data. I still write code, mostly to review what they propose and to decide what ships. If you are building something in health or pharma AI and need someone to own the technology, email me.

Along the way I have been technical lead on an NIH-funded VR and eye-tracking research platform at Texas State University, a research associate at the AUTh Lab of Medical Physics and Digital Innovation on Greek speech recognition, mixed reality and more than 15 European and national research projects, a speaker at more than 15 national events on cloud, and a national winner of NASA Space Apps 2018 and Microsoft Imagine Cup 2014. I write here about once a month on agentic systems, evaluation and teams that ship. The rule I run teams by is simple. Agents propose, deterministic code executes, and a person owns the decision.

Background

  1. Sep 2026 – present

    AI Architect and Technical Lead

    NewPage Solutions, Miami, Florida, US · for a global pharmaceutical client (contract)

    Agentic AI for the client's international commercial and R&D teams, from scoping with senior stakeholders to a validated demo. An AI-driven system supporting a U.S. research laboratory's medicine-formulation workflow, presented to R&D leadership in the first month. Delivery pipelines and standards for the client toolset (Snowflake, Databricks, an enterprise LLM gateway, LangGraph, n8n), LLM cost governance and the data flow between GxP-regulated environments and AI services. Technical direction for the embedded engineering team.

  2. Jun – Aug 2026

    AI Architect

    Proper, Athens, GR (contract)

    Owned the AI architecture: an end-to-end agentic operations layer (LangGraph, n8n, FastAPI) that triages every request across hundreds of buildings and thousands of residents, and helped close the seed round. Set the engineering standard as the senior external AI expert: 14,000+ production messages replayed against a 124-case ground truth and a 24-rule gate before go-live, 94 to 96% action accuracy.

  3. Jan 2025 – Apr 2026

    Chief Technology Officer

    Augmented Training Systems, Austin, Texas, US

    Co-authored the proposals that won non-dilutive federal funding from the Department of Defense and the State Department, and delivered an AI-driven simulation platform deployed to U.S. embassies worldwide, featured in State Magazine in 2025. Built agentic automation (n8n-hosted agents, RAG, MCP) across contracts, invoicing, daily-scrum risk analysis and the ATS Hero platform's engineering and content pipelines. Co-led the zero-to-one journey with the President and the C-level team to a multi-million-dollar business, owning all technical execution. Led 16 people across 3 distributed teams: engineering, 3D, product and operations.

    Press
    Press · 2025Virtual reality transforms crisis training for nurses at American embassies abroad. State Magazine, U.S. Department of State, February 2025.
  4. Jan 2021 – Dec 2024

    Director of Engineering

    Augmented Training Systems, Austin, Texas, US

    Partnered with enterprise and healthcare stakeholders, Kaiser Permanente, Advocate Christ Medical Center and the University of Arizona among them, to turn ambiguous needs into delivered solutions and recurring engagements across 12+ organisations. Led a cross-functional team of 10 (engineers, artists, delivery, proposals and sales) to ship 20 real-time simulation products. Cut delivery cycle time by up to 80% through automation-driven workflows and analysed 100,000+ training actions to improve product and client outcomes. Directed the ATS Hero platform: multiplayer real-time 3D, 50+ interactive IK avatars, real-time body kinematics, low-spec optimisation.

  5. Oct 2023 – Nov 2024

    Technical Lead

    Texas State University, San Marcos, Texas, US · NIH-funded study (contract)

    A secure Azure cloud platform and analytics for an NIH-funded VR and XR study: the metrics and eye-tracking signals captured, role-based access, ethics and data-governance alignment, and a documented API toolkit for 1,000+ users across 3 labs.

  6. 2020 – 2026

    PhD, Biomedical and Medical Engineering

    Aristotle University of Thessaloniki, School of Medicine, Thessaloniki, GR

    MIDAS, at the Department of Transplantation Surgery, Hippokration University Hospital: a modular VR, biosensor and AI platform for surgical training and intra-operative cognitive support. AI on multimodal clinical data to recognise performance patterns and biomarkers, and a voice-driven LLM assistant (RAG, n8n) for surgeons' pre-op preparation and post-op reflection. National fellowship from the Hellenic Foundation for Research and Innovation. The MIDAS page.

    Publications
    Under review · 2026MIDAS: a unified platform architecture for telemetry-triggered Socratic LLM tutoring in virtual reality laparoscopic training and consumer-wearable peri-operative monitoring of surgeons.
    Published · 2024Virtual reality in transitional care: 360º videos for stress management in transition to home care. PETRA 2024, the 17th International Conference on Pervasive Technologies Related to Assistive Environments. doi 10.1145/3652037.3663921
    ›
    Published · 2023Exploring the use of virtual reality in surgical education. World Journal of Transplantation 13(2). doi 10.5500/wjt.v13.i2.36
    ›
  7. Oct 2019 – Dec 2020

    Senior Software Engineer, employee number one

    Augmented Training Systems, Austin, Texas, US

    Built the first real-time training simulations, Mass Casualty Incident and Safety Protocol, and 4+ multi-victim scenarios optimised for low-spec devices. 500+ users globally, piloted in a major hospital.

  8. Jan 2018 – Jan 2023

    Research Associate

    Aristotle University of Thessaloniki, Lab of Medical Physics and Digital Innovation, Thessaloniki, GR (part-time, affiliated)

    Trained a custom deep-learning speech recognition model in Python and TensorFlow, building the full data and training pipeline for an offline assistive device. Contributed to 15+ EU and national R&D projects and microgravity sleep studies, co-authoring winning proposals and 18+ peer-reviewed articles on medical data and AI.

    Publications
    Published · 2024Virtual reality reusable e-resources for clinical skills training: a mixed-methods evaluation. Personal and Ubiquitous Computing 28(3). doi 10.1007/s00779-024-01805-9
    ›
    Published · 2022Feasibility and acceptance of virtual reality reusable e-resources embedded in healthcare curricula. IMET 2022, International Conference on Interactive Media, Smart Systems and Emerging Technologies. doi 10.1109/IMET54801.2022.9929810
    ›
    Published · 2022Errata: From the Guantanamo Bay crash to objective fatigue hazard identification in air transport. SAE International Journal of Aerospace 15(1). doi 10.4271/01-13-02-0017.1
    ›
    Published · 2020Real-time affective measurements in medical education, using virtual and mixed reality. Brain Function Assessment in Learning, BFAL 2020, Lecture Notes in Computer Science. doi 10.1007/978-3-030-60735-7_9
    ›
    Published · 2020From the Guantanamo Bay crash to objective fatigue hazard identification in air transport. SAE International Journal of Aerospace 13(2). doi 10.4271/01-13-02-0017
    ›
    Published · 2020Education, using virtual and mixed reality. Brain Function Assessment in Learning, BFAL 2020, second international conference.
    ›
    PublishedNovel technologies in interactive media: full papers. Conference proceedings volume.
    ›
  9. Jan – Dec 2017

    Senior Software Engineer

    Aristotle University of Thessaloniki, Thessaloniki, GR

    Mixed-reality and augmented-reality medical-education modules (anatomy and physiology) and Android apps, with measurably higher student engagement than traditional methods. CRM web applications (ASP.NET, SQL Server) for healthcare and academic institutions.

  10. Feb 2017 – Aug 2020

    Senior Software Engineer, Technical Orchestrator

    Microsoft Innovation Center, Athens, GR

    Technical Orchestrator on a European Horizon 2020 Parkinson's digital-health study: the full Azure cloud data lifecycle (Blob, Functions, web services) plus AWS S3, securely capturing 30,000+ records from 800 participants across patient, family and healthy-control cohorts via Android. Co-authored 5+ publications and delivered 15+ Azure and cloud talks at innovation events in Greece.

    Publications
    Published · 2020Innovative Parkinson's disease patients' motor skills assessment: the i-PROGNOSIS paradigm. Frontiers in Computer Science 2. doi 10.3389/fcomp.2020.00020
    ›
    Published · 2020Medical follow up assessments of iPrognosis application users for early Parkinson's disease detection. Parkinsonism and Related Disorders 79. doi 10.1016/j.parkreldis.2020.06.155
    ›
    Published · 2017iPrognosis: towards an early detection of Parkinson's disease via a smartphone application. 90th Congress of the German Society of Neurology, DGN, Leipzig.
    ›
  11. 2016 – 2018

    MSc Medical Informatics

    Aristotle University of Thessaloniki, School of Medicine, Thessaloniki, GR · PROMESIP, with Electrical and Computer Engineering and Informatics

    Medical and biomolecular data management, biomedical technology, medical telematics and personalised health, with the programming, signal and image analysis behind them. Thesis: a mixed-reality application on HoloLens that shows the long sensory and motor pathways of the spinal cord and brain as a hologram, and a quasi-randomised trial with 200 second-year medical students at the University of Leeds comparing it with the curriculum's anatomy drawing screencasts, pre- and post-test.

    Publications
    Published · 2019Blueprinting an edu-centric design and development workflow for a prototype mixed reality neuroanatomy resource. iLRN 2019 proceedings, Immersive Learning Research, Academic 1(2).
    ›
  12. 2015

    Digital Health Academy, Digital Healthcare Management

    The University of Sheffield, Sheffield, UK · Startify7, a Horizon 2020 academy

    A European summer academy on digital health entrepreneurship for the hospital and health sector: lean business methods, then business models for digital health, built in a team around a product idea and pitched to practitioners, investors and business angels.
    First prize, Digital Health Academy.

  13. Mar 2015 – Feb 2017

    Software Engineer

    Dataverse, Thessaloniki, GR

    The Greek Labour Inspectorate platform (WPF, IoT and GPS) and Windows, Android and iOS apps across 3 national projects. Contributed to Docutracks (Angular, .NET).

  14. Sep 2013 – Jan 2016

    Co-founder, Software Engineer

    MicroView, Larissa, GR

    A medical-imaging R&D tool automating white-blood-cell detection in lab samples, 87% accuracy in initial lab validation. Two national awards and the 2014 Microsoft Imagine Cup global semifinals.

  15. Sep 2012 – Aug 2014

    Software Engineer

    Department of Computer Science and Telecommunications, Larissa, GR

    Forecasting and decision-support applications for the ARCHIMEDES IV national research program, integrating 500+ environmental and road-condition data points with pattern-recognition algorithms for real-time road-safety prediction.

  16. 2010 – 2015

    BSc Computer Science and Engineering

    Technological Educational Institute of Thessaly, Larissa, GR

    Computer science and engineering: algorithms, mathematics and machine learning, electronics, robotics and networks. Microsoft Student Partner from 2012 to 2016 through the degree.

Work

  1. An AI-driven, end-to-end triage system for a property manager

    Every tenant and owner request, by email or chat, across hundreds of buildings and thousands of residents, read by agents that classify it, extract what matters and propose the action. A 24-rule gate and deterministic code carry it out; a person owns the rules. 14,000 real messages replayed before launch, 94 to 96% action accuracy.

    Proper · 2026
  2. AI on surgeons' fatigue, stress and recovery around real operating days

    A serverless wearable platform (heart rate, HRV, sleep stages, reaction time) with a watchdog for missed wear or sync, and machine-learning models on the multimodal data to recognise fatigue and stress patterns and the biomarkers behind them, operating day by operating day.

    MIDAS, the PhD · 2021 to 2026
  3. ▶

    A VR surgical simulator with a Socratic AI tutor

    The first study: a cholecystectomy in VR with hand tracking at the instrument and phase recognition, and a language-model tutor that reads where the hands are and asks the question a good attending would ask, never the answer first. A voice-driven assistant for pre-op preparation and post-op reflection.

    MIDAS, the PhD · 2020 to 2026
  4. ▶

    Twenty simulations for hospitals, universities and first responders

    Multiplayer simulation with body kinematics, and in the later years agentic automation (n8n-hosted agents, RAG, MCP) across the engineering and content pipelines. 500+ users in the first pilots, delivery time cut by up to 80%, 100,000+ training actions analysed.

    Augmented Training Systems · 2019 to 2026
  5. Crisis training for embassy staff

    An AI-driven VR platform for U.S. embassies worldwide, a multi-year contract won against larger companies and delivered through the State Department's compliance process, its scenarios produced through the platform's agentic pipeline with RAG over the training material. State Magazine, February 2025.

    Augmented Training Systems · State Department program · 2025
  6. ▶

    Tensio Design, a walkable 3D showroom for every catalog

    A side venture. A product catalog becomes a true-to-scale 3D space on the company's own website, no app needed. Early stage, built to learn. tensio.ai

    Tensio Design · 2026
  7. A Parkinson's study across an EU consortium

    Secure capture and lifecycle for three cohorts on Azure, 30,000 records from 800 participants, with partner coordination.

    Microsoft Innovation Center · 2017 to 2020
  8. Greek speech recognition from thirty hours

    Training from scratch failed; transfer learning with a constrained decoder worked, for an offline assistive device that runs the model on the device itself.

    AUTh lab · 2018

Media and events

Academic

Surgeons at the operating table, softened
PhD · MIDAS · Aristotle University School of Medicine · 2020 to 2026

Mixed reality, biosensor and artificial intelligence technologies supporting surgery practice and education

A VR laparoscopic simulator with a Socratic tutor that reads where the hands are and asks the question a good attending would ask, and a wearable study of surgeons' fatigue, stress and recovery around real operating days, with AI on the multimodal data. Department of Transplantation Surgery, Hippokration Hospital.

  • National fellowship, Hellenic Foundation for Research and Innovation
  • Degree due in the first quarter of 2027
  • Unified platform paper under review, 2026
Open the MIDAS page: simulator, data, papers, OSF ›
Publications · 19
  1. 2026
    MIDAS: a unified platform architecture for telemetry-triggered Socratic LLM tutoring in virtual reality laparoscopic training and consumer-wearable peri-operative monitoring of surgeonsUnder review
  2. 2024›
  3. 2024›
  4. 2023›
  5. 2022›
  6. 2022›
  7. 2021›
  8. 2021›
  9. 2020›
  10. 2020›
  11. 2020›
  12. 2020›
  13. 2020›
  14. 2020›
  15. 2019›
  16. 2019›
  17. 2018›
  18. 2017›
  19. —›

Writing

Honours and certifications