CoreStory Taps Upwork To Support Fast-Paced AI Innovation

Roughly two-thirds of enterprise tech budgets go toward maintaining legacy systems — a hidden tax that can impact growth and innovation. As costs mount, more companies are asking whether it’s worth keeping outdated infrastructure afloat.

CoreStory helps organizations chart a path forward. The B2B AI platform is designed to help enterprises understand and modernize massive legacy codebases with a lean internal team and a rising demand for specialized technical skills, especially in AI.

“We build models that are able to understand extremely large amounts of code so that humans can further interact with them on their own terms, which is through natural language,” said Cory Hymel, VP of Product and Research. “A lot of legacy code bases were written 20 to 30 years ago. They still run critical business functions. And they're starting to break down.”

To keep pace, CoreStory blends human expertise with machine intelligence, modernizing code at scale without slowing down.

The challenge: the cost of moving too slowly

Without a B2C user base to generate large datasets, CoreStory runs lean, targeted experiments to validate features and back up market claims.

“In this market, if you're not publishing and coming out with information quickly, you risk falling behind,” said Hymel.

Early on, CoreStory built numerous code products to gather training data on large codebases. Now, as the focus shifts to product development and tuning, the need to move fast has only grown.

That urgency came into sharp focus during a joint research project with Microsoft and GitHub. The study required 300 senior developers with a specific skill set to examine how AI tools are changing the way engineers work.

Hymel knew traditional hiring channels could produce two or three candidates a month, which was far too slow to capture the opportunity.

The solution: speed, flexibility, and a repeatable process

Within six to eight weeks, CoreStory engaged the developers it needed — a process that would typically take a year. “This timeline is generally unheard of in the academic space,” Hymel said. “It gave us more time to interpret the data, put out a better paper, and stay ahead in the market.”

With the breadth and size of the global talent pool on Upwork — made easier by Upwork’s vetting tools, bulk review capabilities, and built-in workflows — CoreStory was able to establish a repeatable process that’s easy to scale.

In fact, Hymel said going from posting to hire can take as little as 15 to 20 minutes. 

“Whether it’s for research studies or just bringing someone in to help build an experiment test harness, the ability to move fast is critical,” he said.

The same process worked for another recent study when the company was able to recruit 30 participants over a week and a half. This time, the company wanted to see how well large language models could identify unexpressed constraints — features or requirements users may expect but not ask for — compared to humans. 

Other studies are in the works. “Being able to run, source, and build experiments quickly is extremely important,” Hymel explained. “There's so much coming so fast. If you're not staying right at the edge of what's going on, you can become irrelevant pretty quick.”

The results: quicker insights, stronger positioning

Speed is a competitive advantage in the tech industry, especially when AI is involved. Quality and logistics need to be considered too. “When running a study, it’s really important to have control over how individuals are working. You don’t want to skew the data in any way,” Hymel explained. “The tools that Upwork already has were really advantageous.” 

Trust features built into the Upwork platform have made the process even more reliable. Hymel particularly appreciates payment protection, standardized contracts, and the transparency of client feedback from previous projects. 

“Upwork feels like a very baked, repeatable, and scalable process to run through; it only takes 15 minutes to show members of my team how to do it,” he said.

The immediate impact of elastic talent

As the use of AI expands, so do AI-related challenges and the skills needed to solve them. “The idea of having a fixed team of full-time engineers working on the same project for years — that model is breaking down,” Hymel said. 

But speed and a lean budget don’t always align, especially for AI talent. “The race for strong and capable AI talent is on and Upwork has helped us stay ahead of the curve and scale the team,” Hymel said. “We’re not Facebook or Google. We’ve got to be a little more scrappy about it.”

That scrappiness means CoreStory’s hiring needs can pivot at a moment’s notice; one week they need a Python developer, the next it’s that same skill set plus deep domain expertise. Hymel said he’s confident that talent on Upwork can deliver.

“Being able to elastically scale to meet our needs is the best model for us,” he explained. “There are a lot of strong, capable individuals who are interested in solving hard problems but don’t want full-time roles. They want to bounce between projects and share their skills fractionally.”

Baking this flexibility into their playbook has become essential. 

“The speed of product development, customer feedback, and expectation cycles has gone up,” Hymel noted. “Hiring needs to move just as fast.” That’s why CoreStory is formalizing frameworks for flexible talent, bringing in the right experts for acute, high-priority problems even if their influence is only short-term. “That’s how we think about using Upwork: surgically bringing in the right people for very specific needs.”