Vocality Health Is Building the Real-Time Medical Translator That Hospitals Desperately Need

The Macro: Medical Translation Is Broken and People Get Hurt

Language barriers in healthcare are not an inconvenience. They kill people. A 2023 study in the Journal of General Internal Medicine found that patients with limited English proficiency had 35% higher rates of adverse events compared to English-speaking patients. Misdiagnosis, wrong dosages, missed allergies, failed follow-up instructions. The consequences compound.

The current solution in most U.S. hospitals is a patchwork of in-person interpreters, phone-based translation services, and sometimes just hoping a bilingual nurse is on shift. In-person interpreters cost hospitals $50-150 per hour and often take 30 minutes or more to arrive. Phone services like LanguageLine work but create awkward three-way conversations where nuance evaporates. Video remote interpreting is better but still requires a human on the other end.

The market is large. The U.S. spends roughly $3.8 billion annually on medical interpretation services. That number only covers the formal market. It doesn’t account for the informal workarounds, the family members pressed into service, or the interactions where no translation happens at all because none is available.

Google Translate handles tourist-grade conversations fine. It does not handle medical terminology reliably. “Chest pain radiating to the left arm with associated diaphoresis” is not the same as “my chest hurts.” The gap between consumer translation and medical-grade translation is enormous, and that gap is where people fall through.

The Micro: Harvard CS Meets Speech AI Research

Vocality Health is building a voice AI system that allows doctors to speak in any language during patient interactions. The goal is to replace human medical translators entirely for routine clinical conversations. Real-time, accurate, compliant with healthcare privacy requirements.

The founding team has the kind of background that makes this bet credible. Brogan McPartland is the co-founder and CEO, a Harvard CS graduate who previously co-founded Fulltrack AI, a computer vision sports app that reached 4 million downloads and became the number one sports app in India. He holds three patents and speaks fluent Hindi. His personal connection to navigating international medical systems is part of what motivated the company.

Vivek Jayaram is the co-founder and CTO, also a Harvard CS graduate, with a PhD focused on generative models. He led computer vision at Second Spectrum, which was acquired for $200 million. His research background is specifically in speech generative models, with over 10 published papers and 6 patents. That’s directly relevant. Building a medical translation system that works in real-time voice is fundamentally a speech generation problem.

They’re part of YC’s Winter 2025 batch. The product targets hospitals and health systems, which means a long sales cycle but high contract values once you’re in.

The technical challenge is real. Medical terminology is specialized, context-dependent, and high-stakes. A translation system for healthcare needs to handle regional dialects, medical jargon that doesn’t map cleanly across languages, and the reality that patients under stress often speak differently than they would in a calm conversation. Consumer translation tools are not built for this.

The Verdict

I think Vocality Health is going after a market that is simultaneously obvious and underserved. Everyone in healthcare knows translation is a problem. The solutions available are expensive, slow, or inaccurate. Voice AI has reached a level of maturity where real-time medical translation is technically feasible for the first time.

The moat, if they can build it, is in medical-specific training data and compliance infrastructure. Any competitor can use Whisper or a similar speech model for general translation. What they can’t easily replicate is a model trained on thousands of real clinical conversations across dozens of language pairs, validated against medical accuracy benchmarks, and wrapped in a HIPAA-compliant deployment that hospitals will actually approve.

The risk is the sales cycle. Selling to hospitals is notoriously slow. Procurement committees, compliance reviews, pilot programs, IT integration requirements. A two-person startup can burn through a lot of runway waiting for a health system to say yes. The counterargument is that staffing shortages for medical interpreters are getting worse, not better. Hospitals are increasingly desperate for alternatives.

In 30 days, I want to know how many pilot programs are running and in which health systems. In 60 days, the question is accuracy. What’s the error rate on medical terminology across the top 10 language pairs? In 90 days, I’d want to see whether any health system has moved from pilot to contract. That’s the signal that separates a demo from a product.