Samacare: AI-Powered Healthcare Platform Optimization

The Problem

Healthcare providers hated Samacare's Chrome extension.

It worked, but it kept interrupting their workflow. Doctors and nurses are already buried in administrative work, and the last thing they need is software that makes their day harder.

Meanwhile, Samacare's own team was drowning in manual data entry. Every prior authorization request meant someone reading through pages of medical PDFs, pulling out information by hand and copying it into forms. It didn't scale, it was expensive, and it was exactly the kind of work AI is good at.

Samacare needed healthcare platform optimization on both fronts before its Series B funding round. Investors would ask hard questions about scalability and technology, and the answers had to hold up.

What We Built

Making the Extension Useful

We rebuilt the Chrome extension from the ground up. Instead of constantly demanding attention, it waits for the moment it can actually help.

The new version:

Appears only when it's needed, based on context
Prefills forms from the current patient, saving clicks and time
Works offline, because clinic internet is famously unreliable
Loads faster, because every second counts in a busy clinic

We added keyboard shortcuts for power users, smart defaults based on usage patterns and real-time validation that catches errors before a form is submitted.

The result was a 115% increase in user engagement. Providers went from tolerating the extension to asking for it.

AI That Works in Practice

Healthcare runs on PDFs: prescription forms, lab results and medical histories. It's all unstructured data that people have to read and process by hand.

We built an LLM-based system to do this automatically. The pipeline takes in messy documents (scanned faxes, PDFs, handwritten notes), runs OCR where needed and then uses AI to extract structured data. It goes beyond simple field mapping to interpret the medical context.

AI isn't perfect, though, especially in healthcare, where mistakes matter. So we built evaluation systems to check accuracy and a human-in-the-loop review step. The AI handles the documents it's confident about; people review the rest.

The system also improved over time, because every correction became a lesson for the next document.

Patient Enrollment Automation

Before, enrolling a patient in a medication program took manual forms, phone calls and faxes. Yes, still faxes in 2024.

We automated enrollment across multiple pharmaceutical programs. That included API integrations with pharmacy benefit managers, automated eligibility checks, document generation and status tracking. Enrollment became faster and less error-prone, and patients got access to their medications sooner.

Architecture for Scale

The platform was still running on a monolith, and for the Series B the team needed to show it could scale.

We designed an event-driven architecture and started implementing it. Services communicate asynchronously, each piece scales independently, and when one part of the system gets hammered, it doesn't take everything else down with it.

We documented the roadmap, trained the team and got buy-in from engineering leadership, so the plan was already turning into real infrastructure by the time investors looked at it.

The Series B Story

Investors don't invest in technology. They invest in businesses that will grow, and they ask hard technical questions to find out whether you can scale.

We supported leadership through technical due diligence with:

Architecture documentation that explained what we built and why
Security and compliance evidence, since HIPAA isn't optional in healthcare
A scalability analysis of the new architecture
An honest assessment of technical debt and how the team would address it

The technical demonstrations showed working AI features rather than promises, and the roadmap was credible because the team had already started executing it.

Samacare closed its Series B, and the technical foundation we built contributed to investor confidence.

The Results

115% increase in user engagement. Providers who barely used the extension started asking for it.

LLM-based automation of data extraction. Document processing that used to depend on people reading PDFs now runs automatically, with people reviewing the cases the model isn't sure about.

Multi-provider patient enrollment automation. Enrollment across pharmaceutical programs moved from forms, calls and faxes to integrated, tracked workflows.

Successful Series B funding support. The technical foundation and working AI features helped show investors the platform could scale.

What We Learned

AI hype vs. AI value. Everyone talks about AI, but few teams use it to solve specific problems. We focused on narrow, measurable tasks. The goal was never to replace doctors; it was to handle the tedious data work they shouldn't have to do.

Performance is a feature. A faster, quieter extension was a big part of why providers started using it. Users notice speed, even when they don't talk about it.

Healthcare requires accuracy over speed. We could have shipped faster with lower accuracy. We didn't: anything the model wasn't sure of went to a person.

Show, don't tell. In due diligence, a working demo beats a slide deck every time. We showed AI extracting real data from real medical forms, which was more convincing than any roadmap slide.

The hard part wasn't the AI. The hard part was making AI useful inside real healthcare workflows, and we did both.

Need help implementing AI in healthcare or optimizing product performance? Let's talk →

See more technical transformations View case studies →