Skip to main content

From writing code to building autonomous agents

AWS Agents Applied Science Manager Purak J. on intent, career pivots, mentorship, and why a PhD wasn’t the prerequisite he thought it was.

Amazon welcomes career pivots, PhDs, and everything in between

Purak J. didn’t follow the expected path into applied science. There was no PhD, no postdoc, no research lab. He started as a 20-year-old software engineering intern in Bangalore. The role of “applied scientist” wasn’t even something he knew existed.

“I was interviewing for a software engineer role,” says Purak. “It was much later that my manager told me about this role within the company, and I got attracted to try it out.”

That manager was software engineering director Mike J. He hired Purak, gave him full ownership of a production project on day one, and ended up shaping the next eight years of Purak’s career.

Leaders who take bets

As an SDE, Purak kept watching the science team work on problems that fascinated him. But more than that, he kept noticing the gap between what the scientists built and what actually made it into production. Models sat on shelves. Systems and science talked past each other.

He decided to close that gap himself. Mike didn’t tell him to get a PhD. He wrote a letter of recommendation for a part-time master’s at Georgia Tech in machine learning, connected him with a mentor on the science team and relocated him from Bangalore to Seattle.

“At that time, a lot of the applied scientists I was seeing were usually PhDs,” says Purak. “I was hesitant. But I think at Amazon, leaders really value intent and bias for action. Intent is usually more valuable and harder to develop than the skill set which can be learned on the job.”

Finding his place in the room

His first science project wasn’t building a model. It was figuring out how to get a model that had been sitting on the shelf into production. He designed a novel A/B testing methodology from scratch. He reviewed the design with Nobel laureate Guido I., an academic research consultant at Amazon.

“I just scheduled a meeting with him to review the design, which was very surprising to have that kind of access,” notes Purak.

The methodology unlocked bidding models that had been blocked for years. Revenue impact: north of $1 billion, rolled out worldwide. Best paper award. A patent. Published externally at AdKDD, a premier workshop on artificial intelligence for advertising.

“All of this was something that initially I really didn’t think I had the expertise to even start working on.”

What he leads now

After eight years in advertising, Purak cold-emailed a science director leading a new initiative called AWS Agents. There was no opening. He had no experience building agents. The director said yes anyway.

As a Bar Raiser, a specially trained interviewer at Amazon who plays a critical role in the hiring process across Amazon, Purak built his own team, prioritizing curiosity over prior agent experience. From there, he and his team took AWS DevOps Agent from zero to launch in six months. It’s a fully autonomous 24/7 operational teammate that monitors incident queues and resolves tickets without human intervention.

“In a span of eight years, I went from being a software engineer intern in Bangalore to leading a team of scientists across New York, San Francisco, and Seattle, supporting a 250-person organization.”

The common thread

The connection between the intern who built an observability tool and the science leader building autonomous on-call agents isn’t a straight line. It’s a pattern: See a problem that excites you, raise your hand, and learn what you don’t know on the job.

“The advice I would give my younger self would be that you’re only limited by your imagination,” says Purak, reflecting on his experience working at Amazon. “There are a lot of opportunities that come by if you just think big.”

“You're only limited by your imagination. There are a lot of opportunities that come by if you just think big."