AKASA Innovation & Technology Culture

Describe a project you’re especially eager to tackle in the new year.

I’m especially excited to build more capable AI for complex healthcare workflows. Over the past year, we’ve made meaningful progress applying advanced ML and LLMs to extract structure from highly variable clinical and operational documents. In the new year, we’re pushing this even further and expanding the system’s ability to reason across multiple documents, understand nuanced context and surface richer, more accurate outputs. This next phase lets us break through some of healthcare’s most persistent administrative bottlenecks and deliver AI that feels more adaptive, more trustworthy and, ultimately, more helpful to the teams who rely on it every day.

What technologies and/or practices is your team leveraging to tackle this project?

We’re combining state-of-the-art language models with carefully engineered retrieval, validation and reasoning pipelines that reflect real-world healthcare workflows. Our team uses modern ML development tooling, including AI-assisted coding environments, to iterate quickly and with confidence. These tools help us test alternative architectures, generate prototypes, and validate assumptions in hours instead of days. We also lean heavily on collaborative engineering practices like tight feedback loops, well-instrumented experiments and continuous evaluation against real encounter data. Together, these technologies and practices allow us to build systems that are not only powerful but also aligned with the accuracy, reliability and transparency standards that healthcare requires.

How does this project tie into larger company goals?

This work sits at the center of AKASA’s broader mission: bringing AI to some of the hardest, most consequential problems in healthcare. By teaching our systems to understand complex clinical information with the nuance of a human expert, we’re opening the door to more accurate clinical documentation and fewer barriers across the care journey.

What types of products or services does your engineering team work on/create? What problem are you solving for customers?

AKASA builds ML systems designed to automate healthcare revenue cycle management, the complex set of processes that connect clinical documentation to billing and reimbursement. The revenue cycle is one of the most complex and error-prone areas in modern healthcare operations, and our goal is to remove friction from administrative processes so providers and staff can dedicate more time to what matters most: delivering excellent patient care.

One of our products enables medical coding. This involves creating systems that analyze clinical documentation from inpatient encounters and automatically assign ICD and PCS code, the standardized diagnostic and procedural codes used across the healthcare industry. These codes form the backbone of everything, from insurance billing and reimbursement to quality reporting and public-health tracking.

Traditionally, this has been a highly manual, detail-heavy task, requiring medical coders to review lengthy clinical notes and translate them into precise codes. It’s not only time-consuming but also susceptible to human error and inconsistencies. 

Tell us about a recent project where your team used AI as a tool. What was it meant to accomplish? How did you use AI to assist?

Our research and engineering team uses a variety of AI-powered developer tools, including Cursor and Claude Code, to enhance our software development workflow. These tools serve as AI pair programmers, helping engineers brainstorm, design and debug code.

Recently, we have been using AI to accelerate experimental velocity, increasing the rate at which we can stand up prototype machine learning training and evaluation pipelines. Rather than spending hours poring over documentation, writing boilerplate code or manually testing small logic changes, working with coding agents allows us to divert more bandwidth to experimental design and analysis. The focus of hands-on-keyboard coding has shifted to regions of the codebase that are highly technical and/or bespoke, those where the internet would be of no help. For a new engineer ramping up on a codebase as complex as ours, having an AI assistant means being able to learn faster and contribute sooner. This approach embodies one of our team’s core principles: Use AI not just in the products we build but also in how we build them. 

What would that project have looked like if you didn't have AI as a tool to use? How has AI changed the way you work, in general?

Without access to AI development tools like Cursor or Claude, each ML experiment would involve significantly more manual effort and incur a longer turnaround time. The absence of AI would also introduce more cognitive overhead and result in higher expenditure of mental energy on mechanical tasks. 

For new team members like me, the onboarding curve would be much steeper, as they’d need to learn both the codebase and the domain context — without the real-time conversational code search and analysis AI now provides. Coding agents also empower those of us on the team who don’t come from traditional computer science backgrounds, broadening the surface area of code we can effectively collaborate on. In essence, AI has become a force multiplier for our teams, in engineering and beyond, helping us move faster, think bigger and deliver more intelligent solutions for our customers.