Medical AI that works in the operating room, not only on a benchmark
FC-OE is a surgeon-led advisory for medical AI. I work with founders, investors, consulting firms and hospitals on one question: will this survive real clinical workflows, real data and real regulators? I am a practising orthopaedic surgeon and medical AI researcher in Zurich, so advisory capacity is deliberately limited.
Areas of work: clinical validation and study design for AI, EU MDR and FDA SaMD readiness, large language models and agents in clinical settings, imaging AI (including fidelity checks on AI-accelerated MRI/CT), wearables and sensor data, movement and biomechanics analysis, surgical and physical AI including robotics, clinical data infrastructure and governance, and funding strategy for AI-driven projects.
Why do medical AI models fail in clinical workflows?
Most hospital data is unstructured and multimodal, and models trained on clean academic datasets break on it. Three failures recur:
- No fit with the workflow. A working repository is not a working operating-room tool.
- Screen dependence. A tool that makes a clinician look at a dashboard, click through screens or type a prompt will not be used. It has to run in the background and surface only what matters.
- Weak validation. Data leakage, bias and unrepresentative test sets fail EU MDR and FDA SaMD review, and they fail clinicians first. Validation has to be designed in from the start.
Algorithms do not treat patients; clinicians do. FC-OE closes the gap between ML engineering and frontline practice.
How do I work with medical AI startups?
Fractional advisory for Seed and Series A MedTech, digital health and clinical AI companies. I stress-test your clinical thesis, validation design, regulatory path and clinical-adoption story. I do not write your code.
- For: Seed and Series A MedTech, digital health, clinical AI.
- Typical questions: Is the endpoint clinically meaningful? Does the model generalise beyond the site that produced the data? Which device class and evidence package fit this product? Would a surgeon actually use it mid-case?
- Model: equity-first for vetted startups. Standard FAST agreement, typically 0.5% to 1% advisory equity with vesting, in exchange for continuous asynchronous oversight. Fee-based work is possible.
How do I advise healthcare investors?
Pre-term-sheet clinical and technical due diligence on MedTech and clinical AI assets.
- For: healthcare VC, private equity, angel syndicates.
- Typical questions: Does the model hold up on noisy real-world data or only on curated sets? Is retrospective bias inflating the results? Can the privacy and security architecture pass a hospital IT audit? Is the product a single-user tool or does it coordinate a whole care team?
- Model: flat fee per asset, or monthly retainer for deal-flow screening and technical memos.
How do I help hospitals and established companies entering medical AI?
Hospitals evaluating AI, and established manufacturers, suppliers and technology companies moving into MedTech or digital health, need the same thing: a clinician's view before committing budget.
- For: university and regional hospitals, health systems, clinical departments, and corporates in materials, devices, sensors, software or pharma-adjacent industries entering the space.
- Typical questions: Where does AI create value in our operating-room, staffing or imaging workflows, and where is it vendor hype? Which clinical problem is worth building for, and who in the hospital must say yes? What evidence, regulatory class and pilot design would a hospital actually accept? How do we run a first pilot with a clinical partner without a long procurement cycle?
- Model: a scoped workshop or opportunity assessment, then retained advisory if useful. A first scoping conversation is free.
Can consulting firms book me as an expert?
Yes. Consulting firms, expert networks and strategy teams can book me for expert calls and short projects on clinical AI, orthopaedics and sports medicine, surgical technology and hospital operations.
- For: Tier-1 and boutique consulting firms, expert networks, corporate strategy and corporate development teams, enterprise clinics.
- Typical topics: the clinical reality behind AI vendor claims, adoption barriers in operating rooms and clinics, reimbursement and regulatory timing, market-mapping sprints, and the shift from EHRs as systems of record to AI that acts on continuous, preventive data.
- Model: hourly calls through your network or directly, or a retained sprint. I do not share patient data or confidential information from any other engagement.
Can I book you for a keynote, workshop or talk?
Yes. I give talks on what AI can and cannot do in medicine today, with evidence graded and vendor claims separated from results. Formats range from a 20-minute impulse talk to a half-day workshop.
- For: hospital boards and management teams, medical societies and congresses, grand rounds, corporate events, university courses, and industry R&D teams.
- Typical topics: AI and operating-room capacity, staffing and clinician workload; large language models and agents in clinical practice; wearables and digital biomarkers; physical AI and robotics in surgery; how to validate and regulate clinical AI.
- Format: in person in Switzerland and the DACH region, or by video worldwide. Talks are informational: options and evidence, not a sales pitch.
- Model: a fee per talk or workshop, plus travel. Send the audience, date and the question your audience is asking.
Start a conversation
Email felix.oettl@fc-oe.com. For advisory, include your technical documentation, Clinical Evaluation Plan (CEP) or investment memo. For talks, include the event, audience and date. For expert calls, include the topic and the questions you want answered. Advisory mandates are capped, so a document-backed first message gets the fastest answer.