Pursuit hosted Raj Mehta and Anushka Gupta of Percepta, an AI transformation company launched in 2025, for a fireside chat with our Builders. Both Raj and Anushka came to Percepta from traditional software engineering roles. Their new title, Applied AI Engineer, barely existed when they started conversations with Percepta. As we wrote about recently, AI model companies and traditional consulting firms are racing to define this role under names like AI Forward Deployed Engineer. This is exactly the kind of role that our Builders are taking on for businesses and nonprofits.
Here's the recap.
Percepta works with customers in essential industries like healthcare, civil services, supply chain and manufacturing, and financial services to transform their organizations with AI. Their teams of researchers, engineers, and product managers embed inside organizations to build and ship custom solutions.
OpenAI and Anthropic have launched similar consultancies, but Percepta’s model is different. They focus on embedding deeply within their clients’ organizations to understand what they might need, rather than recommending a specific product.
As Raj said:
"Sometimes the solution might be as simple as helping teams integrate Claude or ChatGPT so that they can build their own things. Other times it might be building out a custom workflow for them, sometimes training open source models for specialized intelligence or data privacy concerns."
Raj and Anushka call themselves engineers, not consultants, because they ship. But what they ship actually benefits the organizations they work with, because most of the work happens before they write code.
As Anushka said:
"My role involves a lot of exploration. Day to day, my role might be talking to doctors, talking to nurses, seeing what they do, mapping out a patient as they go in the hospital. Or it might be, just sitting down and coding. So it's a little PM work, some engineering work, but really just understanding the whole realm of AI opportunities."
Both Raj and Anushka have been working for the same client: a large hospital system in the midwest, but on different teams. One problem that was flagged immediately was the nurse scheduling system, which was still being created manually by hand, a tedious exercise that took hours every week. Rather than building something fast, the Percepta team stopped to learn more. They spent multiple days inside different nursing units, sitting with the nurses and managers who actually built and lived the schedules. They asked questions, observed, and shadowed enough people to figure out which challenges were unique cases and which ones repeated.
The obvious one: no back-to-back shifts and no too many shifts in a week, both for burnout. The harder ones: each nurse had limits that shifted by the week, sometimes requiring a full reshuffle on short notice. Compatibilities mattered too, between nurses and with specific doctors. The nurse manager had been scheduling by pencil for years to keep all these nuances in play. As Raj put it, there were many things the team did not even know had to be taken into account.
Anushka added in:
"If you look at someone's entire workflow day to day, just really shadow them, understand what they're doing, you might understand that there are different solutions to their problems. But the only way that you're able to do that is by building a relationship with them, meeting them multiple times, chatting over coffee, just understand what their problems are and then come back with holistic solutions."
Raj called scoping the hardest part of the work:
"Scoping is sometimes the most challenging thing to do when you approach a client because everyone thinks that they need AI and they think that it'll solve all their problems. But AI isn't always the right solution."
A traditional engineer takes tickets from a product manager, builds them, and waits for feedback. The Applied AI Engineer role works differently.
As Raj said:
"Nowadays, building software is becoming easier and easier. The real challenge now is not can I create pixels on a screen, but really consider: what is the problem, and what should I make? What should I make to solve it is the skill that we are trying to build."
Pursuit's Builders recognize this firsthand. The first part of their industry-facing project is writing a strong product requirements document. They're building instincts about what to build and why, not just how to build it.
Anushka added a discipline that's easy to miss in scoping: making sure the solution works across users.
As she said:
"You won't just have one user, you're going to have multiple people doing the same job. So see what is actually replicable and long-lasting."
Finally, meet users where they are. With clients who fear AI, frame the solution in their own vocabulary. Whether that's ChatGPT or simply "technology," name what the solution will actually do for them.
Raj was direct about what happens after you ship:
"You might have thought, okay, I built the software, they're gonna use it in this way, and then you hand it off, and then they use it completely differently than you expected."
Anushka added a related point. When you have been heads-down on a project for two months, the version in your head is not the version a first-time user actually opens. The instructions matter. The defaults matter. User design matters more than you think it will when you are deep in the build.
And then there are the nights the code breaks. One night the tool ran and a group of nurses got no shifts for the entire month. The nurse manager called immediately. The Percepta team listened, opened the code, traced the issue, fixed it, and apologized.
Anushka was clear about the standard:
"If it's a P0, if it's 9 p.m., you're working on it. You're responsible for a product that the users are using. Showing that willingness to help them when something does go wrong helps build trust and long-term success."
What Percepta is really shipping, in other words, is the willingness to be on the other end of the phone when a tool breaks.
Raj answered directly: most of the time, the leap from "could be automated" to "actually automated" never happens. Workflows have nuances that don't survive the leap. His example: an insurance claims processor handling cases one at a time. An AI system clears the routine backlog while the human takes the hard ones. More meaningful work, same headcount, less overtime.
The wider context, Raj said, is that the headlines about replacement do not match what he and Anushka are doing on the ground:
"There's still the gap between what's happening at the frontier of tech and what companies across all industries want, and that gap is just getting larger and larger."
The nurse manager still holds constraints in her head that the scheduler will not capture. The job is to give her a better tool. That is the actual work emerging across applied AI engagements, and it is a different role than the one the headlines describe. It is quieter. It is more iterative. It looks more like fieldwork than like ML. And it is going to need a lot more practitioners to close the gap.
Raj and Anushka had to spend weeks learning how a hospital actually runs. They went unit by unit, asked questions, took notes. Most of what they learned was not about software. It was about how a nurse manager balances her constraints, what an aging printer is or is not connected to, what a stressed-out person will and will not tell a stranger honestly.
Many Pursuit Builders walk into our program with lived experience in those kinds of things. They have worked in hospitals, in shelters, in case management, in classroom support, in city services, in frontline retail and logistics. They know what an underfunded ops team cannot get to. They know which constraints never make it into a brief. Pursuit's AI-Native Program layers applied AI fluency on top of that lived experience.
Our Builders are doing this work right now: inside small and medium-sized businesses through our Amazon and Google pilots, inside nonprofits through the AI Nonprofit Build Corps with Mizuho and Goldman Sachs, across capstone projects with VC, PE, and startup partners, and now in full-time roles as well.
Anushka closed with the line that clearly summarized the future potential:
"This job will be on the rise, and you guys are getting in on it very early on."
