Transforming diagnostic test results into actionable summaries with AI
Context
Turi Saúde is an early-stage healthcare startup building AI-assisted workflows to reduce physicians’ administrative burden.
Reviewing diagnostic test results is a major source of friction, which Turi aimed to automate.
Team Role
Owned UX strategy and end-to-end design, partnering with the CEO on product and business decisions and with engineers on technical constraints. Built and validated the platform through vibe-coding, analyzing usage and cost data to inform prioritization.
Outcome and Impact
🧪
Product Outcome
Launched Turi Saúde Lab, a freemium platform offering AI-powered tools to automate manual, repetitive workflows.
❤️
User Engagement
5K+ diagnostic test results processed
28.4% of monthly usage rate
NPS: 9.25
📈
Business Impact
~33% of inbound leads generated
B2B partnership negotiations with a multinational pharmaceutical company and healthcare startups
Problem framing
Lab tests and imaging exams are a core part of clinical diagnosis, yet in Brazil results often reach physicians in lengthy, fragmented, non-standardized formats that require manual review, data extraction, and documentation in electronic health records (EHRs).
The inefficient workflow disrupts clinical care, reduces physician–patient interaction during appointments, and shifts work to unpaid hours, increasing time spent on administrative tasks instead of patient care.
Interviewed Radiologist
[...] it’s always a moment in the consultation when, in theory, you need to be focused because you’re analyzing the exams. But it creates an awkward situation where the patient is waiting. There’s silence and anxiety, because they want to know the results, and you’re looking at the exams instead of focusing on them.
Interviewed Geriatrician
During the appointment, they bring everything printed. So I take pictures and write things down later, otherwise I lose too much time.
Hypothesis
By summarizing diagnostic test results with AI, physicians spend less time on documentation, enabling more attentive, patient-centered care.
Build Process
After defining the high-fidelity designs in Figma and building automations in n8n, I moved into the implementation phase, developing the platform on Lovable. Named Turi Saúde Lab, the platform hosts multiple tools for rapid validation while maintaining a flexible architecture for iteration.
The first tool, Clinical Exam Summarizer, focused on structuring multi-page PDFs into synthesized, ready-to-use clinical annotations with one click.
Initial Launch
After two months of development, I launched the tool to validate the core hypothesis and generate company leads. Users can upload up to three PDF files, which are summarized by AI into concise, factual outputs. With one click, multi-page files are ready to copy and paste into EHRs, reducing manual steps and enabling faster analysis.
Supporting More Files and Image Formats
User surveys indicated strong adoption, reflected in an NPS of 9.25, and highlighted physicians’ need to process more files, including image-based inputs. In response, I increased upload limits and enabled simultaneous processing of PDF and PNG/JPEG files.
Improving OCR Accuracy and Summary Consistency
User feedback revealed incomplete clinical summaries. After testing and desk research, I replaced the OCR tool and refined the AI system prompt, improving extraction accuracy and summary consistency.
Highlighting Abnormalities for Actionable Insights
Physicians’ needs varied by specialty and documentation preferences. To keep outputs actionable, I added a dedicated section highlighting clinical abnormalities.
Freemium Model
Improving OCR accuracy raised per-page costs by 11×, materially impacting unit economics. After validating user interest, Turi Saúde Lab transitioned to a freemium model—capped free usage with unlimited paid plans—to better align revenue with operating costs.
Product Impact and Early Results
Turi Saúde Lab demonstrated strong early traction, processing over 5,000 diagnostic test results by physicians, driving 28.4% of monthly usage rate, achieving an NPS of 9.25, and generating 33% of inbound leads, while enabling early B2B partnership discussions.
Key Learnings and Tradeoffs
Managing two products in one squad required strong prioritization and systems thinking.
Choosing vibe-coding over a traditional development process reduced dependency and opportunity cost.
Prioritizing accuracy over cost ensured output quality—critical in healthcare.
Thank you for reading!
Let's connect!