Dani Kiyasseh
Founder & CEO at Halsted AI
Leading Halsted AI through Techstars '25 to automate surgical documentation and map surgical steps via vision-language models.
Intro
I'm Dani Kiyasseh, the Founder and CEO of Halsted AI. My career has been dedicated to the intersection of machine learning and healthcare, moving from a PhD at Oxford and a postdoc at Caltech to building frontier AI for the operating room. At Halsted AI, we are focused on surgical intelligence—specifically automating operative notes and mapping surgical workflows using nimble, deployment-ready vision-language models. I’m passionate about surgeon-led innovation and ensuring that the tools we build actually survive the 'maelstrom' of real-world surgery. I’m currently looking to connect with surgery residents for our fellowship program and health systems interested in the future of surgical documentation. Whether you're a clinician or a technologist, I'm always open to discussing how we can achieve true surgical mastery through better data and more efficient AI.
Networking
What I can offer
- ›Expertise in surgical AI and vision-language models
- ›Access to the Halsted platform for practicing surgeons
- ›Insights into bridging clinical practice and AI development
What I'm looking for
- ›Surgery residents and fellows for the Halsted AI Fellowship
- ›Partnerships with health systems and payers
- ›Engagement with the digital surgery community
Best fit for
Focus
Current interests
Core competencies
Background
Career
Transitioned from deep academic research at Oxford and Caltech into industry roles at Vicarious Surgical and Cedars-Sinai before founding Halsted AI.
Education
Postdoctoral Fellowship at Caltech (2021–2022); PhD in Machine Learning/Biomedical Engineering from University of Oxford (2018–2021); BS in Biomedical Engineering from Johns Hopkins University (2014–2018).
Achievements
- ›Named to Forbes 30 Under 30 (MENA) and MIT Innovators Under 35 (MENA)
- ›Featured on the front cover of Nature Biomedical Engineering
- ›Developed Halsted VLM, a surgical vision-language model covering 8 specialties
- ›Built an annotated surgical video library of 1M+ videos
Opinions
- Bigger is not better: lightweight models can outperform billion-parameter models in specialized fields.
- Surgical AI often fails because it is evaluated on controlled steps rather than the 'maelstrom' of real surgery.
- Surgeons, not technologists, should dictate the tools used in the operating room.
- Specialty-specific models rarely make economic sense; models should be specialty-agnostic.
Personality
Communication style
Professional, authoritative, and mission-driven, balancing academic rigor with entrepreneurial urgency.
Formality — 7/10
Vocabulary