Product vision, 0-to-1 build, and AI/data-driven decision-making in one of the highest-stakes environments there is.
Every minute of miscommunication in an operating room has a ripple effect: delayed handoffs, idle OR time, and added stress on already-stretched staff. Left unaddressed, this kind of friction doesn't just slow procedures down, it erodes staff morale and retention in one of the highest-turnover, highest-stakes environments in any hospital.
Hospital leadership needed evidence, not anecdotes, that a technology intervention could actually move the needle on efficiency and staff experience, before committing real budget to build it.
I defined the product vision and roadmap, then ran user research with 100+ hospital staff to surface the real friction points behind OR inefficiency, not just the assumed ones. I pitched the concept to hospital administrators and secured $10,000 in funding to build an MVP.
I led a team of 4 through the design and build of an AI-powered decision-support product using RASA, leveraging NLP to support real-time communication between surgeons and OR staff across pre-op, intra-op, and post-op stages, with human-in-the-loop escalation for high-stakes moments. In parallel, I analyzed data from 250+ surgeries to identify the factors most affecting OR efficiency and built a dashboard that improved OR efficiency by 25% and modeled an estimated $9K+ in daily savings from efficiency gains, giving hospital leadership real-time, decision-ready insight.
The central design tension: more automation could reduce communication burden faster, but risked clinician trust if it felt like it was making judgment calls it shouldn't.
| Option | Staff Burden | Clinician Trust | Build Complexity | Verdict |
|---|---|---|---|---|
| Fully Automated Status Narration | Lowest | Low, feels like a black box | High | Rejected |
| Manual Status Boards (no AI) | Highest | High | Low | Rejected, doesn't scale |
| NLP Assistant, Human-Confirmed Signals | Moderate | High, transparent & bounded | Moderate, buildable by a 4-person team | Selected |
We scoped the RASA-based NLP assistant to surface case-status signals conversationally, always verifiable by staff, with human-in-the-loop escalation built in for high-stakes moments - giving a fast, buildable MVP that clinicians could actually trust in the room.
100+ hospital staff interviews to map real communication breakdowns across surgical teams.
Packaged the product vision and secured $10,000 in funding from hospital administrators to build the MVP.
Analyzed 250+ surgeries to identify the highest-leverage factors driving OR efficiency.
Led a team of 4 to build the RASA-powered NLP assistant MVP, with guardrails and human-in-the-loop escalation, integrating with existing hospital workflows.
Shipped a real-time efficiency dashboard adopted by hospital leadership alongside the AI assistant rollout.
A funded MVP starts with evidence. 100+ interviews and 250+ surgeries of data gave the pitch credibility that opinions alone never would have.
In high-stakes environments, the right scope for AI is often narrower than it's technically capable of, and that's a feature, not a limitation.
Efficiency and cost savings are lagging indicators of trust. Staff adopted the tool because it respected their workflow and kept a human in the loop, not because it was novel.
