Case Study: Picwell

Customers make wise health care choices using Promptworks.

Picwell: Pick the most cost-effective insurance plan using machine learning

Since 2013, Picwell’s mission has been to help consumers navigate the ever-evolving health insurance industry by recommending plans based on historical claims data.

Picwell came to Promptworks seeking our technical expertise to help them flesh out their existing product, as well as guidance about how and when to increase their engineering capacity to support this updated product.

Here was our challenge:

Picwell didn’t have in-house expertise in statistical modeling, automated data pipelines, or API development to expand on their core product.

The Picwell team also sought to land new clients and prove their value to the market, which created another hurdle: How would they add more skilled developer capacity to support these additional customers?

Here's how we solved it:

Promptworks was able to help Picwell acquire the caliber of talent they needed. Our engineers specifically tackled the following:

1. Improved the data model.
We helped them build a custom, fully-automated data pipeline that ran simulations of healthcare costs across a distributed computer cluster. It then trained statistical models using the resulting simulation.

2. Built a robust API.
Picwell’s clients depended on their API to provide timely and accurate responses. Using Python and Flask, Promptworks engineers helped build a robust API—with an exhaustive suite of end-to-end integration tests—that Picwell’s clients could rely on.

3. Brought Agile methods to Picwell.
Our help in implementing an Agile process at Picwell ensured that we were building the right software and that it would be delivered on-time and on-budget. Picwell saw progress every day and could prioritize features in real time.

See the results:

Since Promptworks’ partnership with Picwell, they’ve raised $10M+ in outside funding and added dozens of new team members.

We were also able to reduce the time to build new statistical models from terabytes of data; what formerly took weeks could now be done in a few days. By adding automated testing and quality analysis around the model creation process, we were able to increase confidence in the end result.

By maintaining a suite of integration and load tests, we also ensured that the API was widely available, especially during peak traffic.

We'd love to keep in touch!

Opt in for occasional updates from us. Privacy Policy