How Thatch is Building the Flexible Future of Healthcare Benefits on Render

Industry

Health

Region

North America

Thousands of businesses rely on innovative AI healthcare startup Thatch to provide flexible health benefits. Thatch always aims to be at the forefront of the industry, and their customers also expect their platform to be reliable, secure, and compliant. In turn, Thatch’s engineers rely on Render to ship features faster while also enabling HIPAA compliance and reliability at all times.

With Render, Thatch’s engineering lead Kristina Shia found a solution that provided HIPAA and SOC2 compliance without compromising developer velocity.

“When HIPAA compliance became a priority for us, we didn’t have to shift focus away from Product. Render handled the security building blocks like private networking, data encryption, and audit controls so our engineers could keep moving fast without taking on compliance overhead.”

Kristina Shia, Director of Engineering at Thatch

We’ve asked Kristina about her experience using Render. In this interview, you will learn:

What Thatch is and how their engineers ship AI and revenue-generating features faster with Render
How Thatch prioritizes speed versus thoroughness in their engineering processes
Key components to Thatch’s stack that have helped them grow their customer base and offerings.

Health insurance choice is here—it’s just not efficiently distributed

Render: Tell us about your mission at Thatch.

Kristina: Thatch is a new way to do healthcare, and our mission is to help businesses give really great health benefits to their team.

Using Thatch, a company sets a healthcare budget for each employee. Then each employee can choose how they want to spend it. For example, someone might select a cheaper insurance plan with a higher deductible, and use their remaining budget toward other medical expenses like braces for their child. This flexibility also means that someone with a chronic condition who needs to see a specific provider can always choose a plan that their doctor accepts. We’ve seen employees reduce the cost of their weekly therapy sessions from $150 to $15, just by choosing a plan so their therapist is in network.

Employees can choose their plan and manage healthcare expenses directly on Thatch

In contrast, if you work at a startup or SMB today, you might be used to picking between just one or two healthcare plans decided by your employer, if you have health insurance at all.

If this mission speaks to you, Thatch is hiring!

Solving engineering problems with a healthy dose of AI

Render: What are some unique engineering problems you get to solve?

Kristina: Great question! There’s a surprising amount of greenfield technology to build in our space. The legislation that enables our product, called the Individual Coverage Health Reimbursement Act or ICHRA, was only passed in 2020. Every week we learn something creative about how we can build the tech and funds flows to leverage it.

For example, a few months ago we were brainstorming ways to help employers decide on their healthcare budget. It’s tricky because healthcare costs vary across the US. We realized that since we had the cost data, we could build a model for them to simplify that decision.

As another example, we’ve built and deployed AI assistants to help our customers navigate healthcare choices. Today, a Thatch user can ask our chatbot what a deductible is, and can then get help selecting an appropriate insurance plan from all the options in the marketplace. When a Thatch user is at the grocery store, they can also text the chatbot to ask, for example, if Ibuprofen is an eligible expense. We’ve been careful to deploy AI only in areas where we’re confident it will work well.

An AI agent helps employees choose from dozens of health, dental, and vision insurance plans.

Our customers span many sectors, including nonprofits, retail, trucking companies, and AI startups. As we build, we think carefully about serving such a wide range of audiences.

Render: Tell us about the demo app you’ve built for the Thatch sales team.

Kristina: The demo app is a full-featured version of our production app, with a few modifications. For example, we autofill a few forms to streamline the flow for our demos. This demo app was so easy to make, I almost didn’t mention it. It was simple to deploy on Render:

We use one Render project with three environments: production, staging, and demo.

We deploy the demo app to the demo environment to ensure that it won’t affect our production data. In the demo environment, we set environment variables that point to only test databases and other test-only resources.
Instead of auto-deploying each commit to the demo environment, we manually deploy it once every week. This means our sales teammates aren't surprised by any changes while they’re meeting with a customer. Instead, they can review all changes at the end of each week.

Thatch philosophy: “Make haste, slowly”

Render: You’ve described yourself as a company of seasoned professionals now building a startup. How is that reflected in your engineering practices?

Kristina: One of our principles at Thatch is “Festina lente”—a Latin phrase that translates to “make haste, slowly.” It means we balance the need to meet a deadline with doing things well.

Our experience provides us with a solid intuition about when to go fast, and when to take our time. For example, whenever we make changes to the core data model, we write a doc, shop it around, and do an informal design review. We make sure we’re always building for forward compatibility.

In contrast, we took a ship-fast-and-iterate approach with a quoting tool to help companies estimate costs, because that didn’t affect our core data models.

Thatch uses data from 60,000 health plans to define custom healthcare budgets for each employee.

Render and Rails—two pillars of the Thatch stack

Render: Which parts of your engineering stack have played out as long-term wins?

Kristina: I’ll talk about two things.

The first is Render, quite honestly. Here are just a few bright spots:

We spend few cycles on keeping the app running: Among our ten engineers, two of them are more interested in building platform- and security-shaped work. Thanks to Render, they can focus much more on problems specific to Thatch.
Render is accessible to the whole team: Any engineer can go into Render and understand if they need to update a specific config. There’s basically no learning curve.
Seamless GitHub and CI/CD integration: We’ve saved a lot of cycles by not having to stand up any of the CI/CD pipeline. Today at Thatch, it takes 6-8 minutes to go from a committed PR to live in prod.
It's easy and fast to deploy new apps and internal tools: For example, using Render, we were able to quickly spin up our own hosted version of Metabase. It took one engineer a half day to set up. We’ve plugged Metabase into a PostgreSQL read replica and Redshift, which stores our marketing data. Metabase has empowered everyone at the company to dig into data questions like: “Where did this customer come from? Did they come in as a lead from sales?”, and Metabase links are now often passed around in our conversations in Slack. In terms of effort to payoff ratio, this was a large asymmetrical win.

Second, we’re very pro Ruby on Rails, which might be a mildly controversial take. It’s been a big source of leverage for us.

Because the Rails ecosystem is so mature, the features we want are usually built in, and if not we can find high-quality libraries. For example, we’ve used third-party gems to implement an async job execution framework, audit logging for database models, and other functionality that work great out of the box, which we otherwise would’ve needed to invest in in-house.

Together, Render and Rails help us stay focused on solving core problems for our users.

Render: Thanks so much Kristina! Where can folks find you if they want to get in touch or learn more about Thatch?

Kristina: You can find me on LinkedIn or send us a note at hello@thatch.ai. We’re also actively hiring engineers who are excited to shape the future of healthcare.