UniHealth Blog

Insights, stories, and updates from our journey building the future of healthcare enrollment

Featured Articles

The Scam Problem: Why Medicare Open Enrollment Is a Nightmare for Older Adults

During Medicare open enrollment, older adults get upwards of 50 calls per day from agents trying to sell them Medicare plans. When you input your phone number into an online enrollment website, what you don't realize is that your phone number is sold, then sold again, passing through many hands, often illegally. This is the reality millions of older Americans face every year.

Founder•January 18, 2025

Company Story12 min

The Origin Story of UniHealth: From Medicine to Mission

I grew up in China, and when I was 11, we moved to Vancouver, Canada. Throughout high school, I was laser-focused on medicine. I wanted it so badly that two of my childhood best friends followed the same path—one became a doctor, another became a nurse. Then I got into UC Berkeley as an international student. This is the story of how I went from wanting to be a doctor to building AI for healthcare.

Founder•January 15, 2025

Technical Deep Dive20 min

Why UniHealth's Constraint-Based Approach Revolutionizes Medicare Plan Recommendations

Traditional Medicare recommendation systems rely on heuristics, not mathematics. They can't prove their recommendations are optimal. They can't quantify risk. They can't explain their reasoning. UniHealth's CSP methodology combines formal constraint satisfaction, probabilistic cost modeling, and risk-aware optimization to provide mathematically rigorous, provably optimal recommendations with full cost distributions and natural language explanations.

UniHealth Engineering Team•December 9, 2025

Healthcare Insights18 min

How to Shop for Insurance Agents: 10 Questions That Will Change How You Choose

Most people don't know how to evaluate insurance agents. They ask the wrong questions. They focus on the wrong things. After years of working in Medicare enrollment and building AI systems that outperform human agents, here are 10 questions that will help you find an agent—or service—that will actually serve your best interests. Does your agent answer questions 24/7/365 in 90+ languages? Can they guarantee their recommendations are unbiased and mathematically optimal? Do they simulate all risk scenarios?

UniHealth Team•December 9, 2025

All Articles

Technical Deep Dive15 min

Why AI Agents Are the Only Way to Solve Medicare Enrollment

When I first started building UniHealth, I thought Medicare enrollment would be straightforward. Build a database of plans, create some filters, add a comparison tool. Done. I was wrong. After years of building, I've learned something fundamental: Human insurance agents cannot solve Medicare enrollment. Only AI agents can. Here's why.

January 12, 2025Read →

Technical Deep Dive10 min

From 4,000 Columns to One Recommendation

When I first saw a Medicare Advantage plan data file, I was overwhelmed. 4,000 columns. Per plan. Thousands of plans. How do you turn that into a simple recommendation? This is the story of how we built a system that takes 4,000 columns of complex data and turns it into one clear answer: 'This is the best plan for you, and here's why.'

January 10, 2025Read →

Technical Deep Dive12 min

The Provider Directory Problem: 925,077 NPIs and Counting

When I first started building UniHealth, I thought provider networks would be straightforward. A plan has a list of doctors. Check if your doctor is in the list. Done. I was wrong. 925,077 provider NPIs. 470 networks. 210,188 plan-network relationships. This isn't a list. This is a three-layer hierarchy that makes provider networks one of the most complex parts of healthcare.

January 8, 2025Read →

Building UniHealth14 min

What I Learned in 2022: The Year That Changed Everything

2022 was the year I went from knowing nothing about U.S. healthcare to understanding it deeply enough to build a company around it. It was the year of immersion, of deep learning, and of the epiphany that would define everything that came after. This is what I learned.

January 5, 2025Read →

Building UniHealth11 min

Building UniHealth: A Four-Year Journey from Idea to Reality

In my last post, I shared the origin story of UniHealth—how I went from wanting to be a doctor to building AI for healthcare. But that was just the beginning. Today, I want to share the actual journey of building it: the timeline, the breakthroughs, the setbacks, and what we've learned along the way.

January 3, 2025Read →

Technical Deep Dive14 min

The Formulary Challenge: When Your Medication Isn't Just a Medication

You think choosing a Medicare plan is about finding one that covers your medications. Simple, right? Wrong. Your medication isn't just a medication. It's a tier structure. It's a prior authorization requirement. It's a step therapy protocol. It's a pharmacy network nuance. It's a quantity limit. It's a restriction. And if you get any of these wrong, you could end up paying hundreds or thousands of dollars more than you expected—or worse, not getting your medication at all.

January 20, 2025Read →

Healthcare Insights12 min

Why 40% of Medicare Counselors Get It Wrong (And How We're Fixing It)

When I volunteered as a HICAP counselor, I thought I was helping people. I was trained. I was certified. I was doing my best. But here's the uncomfortable truth: Even the best-intentioned, most well-trained Medicare counselors get it wrong 40% of the time. This isn't a criticism of HICAP counselors. They're volunteers. They're doing their best. But the problem is too complex for humans to solve consistently.

January 22, 2025Read →

Healthcare Insights15 min

The Lifetime Penalty Problem: When a Simple Mistake Costs You Forever

As far as I know, no other country in the world has lifetime penalties for late enrollment in their national health insurance program. Only the United States. If you don't enroll in Medicare Part B or Part D when you're first eligible, you pay a penalty—for the rest of your life. Every month. Forever. And most people don't even know this penalty exists until it's too late. This is the story of how veterans who served our country, and countless others, fall through the cracks.

January 25, 2025Read →

Building UniHealth13 min

What I Learned as a Senior Care Commissioner of Santa Clara County

When I was appointed as a Senior Care Commissioner of Santa Clara County, I thought I understood the problems older adults face. I had volunteered as a HICAP counselor. I had worked as a Long-Term Care Ombudsman. But sitting on the Commission, listening to community members, reviewing policies, and seeing the system from the inside—that's when I really understood what seniors are up against. This is what I learned, and how it shaped UniHealth.

January 28, 2025Read →

Building UniHealth14 min

Building in Public: The Community Roles That Shaped UniHealth

Most startups build in private. They work in isolation. They build what they think people want. I took a different approach. I built UniHealth while serving the community. I took on roles that kept me connected to the people I was building for. I listened. I learned. I built based on what I saw, not what I assumed. This is the story of how those community roles shaped UniHealth—and why building in public is the only way to build something that actually works.

January 30, 2025Read →

Technical Deep Dive12 min

Evals as the North Star: How We Measure Success Against Humans and Machines

Most startups measure success by metrics that don't matter: user growth, revenue, engagement. We measure success differently. We measure against humans. We measure against machines. We measure against incumbents. We measure what actually matters: accuracy, completeness, and outcomes. This is our evaluation framework. This is our north star. This is how we know we're winning.

February 1, 2025Read →

Technical Deep Dive14 min

Building a 50GB Healthcare Data Infrastructure

When I first started building UniHealth, I thought the hard part would be the AI. The machine learning. The natural language processing. The agentic reasoning. I was wrong. The hard part was the data. Not just getting the data. Not just storing the data. But building an infrastructure that could handle 50GB of Medicare data that changes quarterly, comes from dozens of sources, and needs to be queried in milliseconds. This is the story of how we built it.

February 3, 2025Read →

Industry Analysis15 min

The Incentive Problem: Why Healthcare Costs 20% of GDP

After reading 50 books and taking 40 Coursera classes on U.S. healthcare, after researching hundreds of startups, after mapping out the entire ecosystem, I came to a single conclusion: U.S. healthcare costs 20% of GDP—roughly double what other OECD countries pay—while delivering lower quality of care. The problem wasn't technology. It wasn't regulation. It wasn't even complexity. The problem was a single word: incentive. And here's the thing: That actually makes it easier to solve.

February 5, 2025Read →

Technical Deep Dive14 min

The AHIP 100%: How Our AI Agent Beat the Industry Standard

When we first built our AI agent for Medicare enrollment, I knew we had something special. But I needed proof. Not just proof that it worked. Proof that it was better than humans. So I did what any insurance agent has to do: I put it through the AHIP certification exam. Our AI agent scored 100%. Not 90%. Not 95%. 100%. This isn't a story about AI being smarter than humans. This is a story about what's possible when you can access all the data instantly, never forget a detail, and reason consistently.

February 7, 2025Read →

Building UniHealth13 min

From Rocket Fuel to Healthcare: What I Learned About Scale

When I tell people I worked at Rocket Fuel, most of them have never heard of it. That's fine. But here's what they're missing: In 2014, Rocket Fuel was one of the top three AI companies in the world. The AI trained over 90 petabytes of data. It placed 250 billion bids per day. It executed with less than 100 milliseconds of latency. I stayed there for five years, working my way up from analyst to leading the data science team. But here's what I learned: The scale thinking I developed at Rocket Fuel applies to healthcare in ways I never expected.

February 9, 2025Read →

Building UniHealth12 min

How We're Different: Why We're Not Another Point Solution

When I first started researching healthcare startups in 2022, I saw hundreds of companies building point solutions. Each one solved a specific problem. Each one added another tool to an already complex ecosystem. I wasn't interested in yet another point solution that made the entire ecosystem even more convoluted versus less. What I wanted to understand was to map out the entire ecosystem and understand, from an ecosystem level, how to actually resolve it. The result: We're not building a point solution. We're building infrastructure.

February 11, 2025Read →