Intern of the Week: Shubhashis Roy Dipta

Name: Shubhashis Roy Dipta

 

Pronouns: He/him

 

Semester of Internship: Summer 2026

 

Major: Computer Science (Ph.D.)

 

Current Class Level: Doctoral Student

 

This semester I completed a(n)...: Internship

 

Internship, Co-op, or Research Site (Company/Organization Name): Amazon (Alexa AI)

 

Location of the Organization (City, State): Bellevue, WA

 

Title of position: Applied Scientist Intern

 

Tell us about your internship, co-op, or research opportunity, including your day-to-day responsibilities: 

I'm an Applied Scientist Intern at Amazon Alexa AI, where I research self-distillation with reinforcement learning to improve LLM reasoning on agentic tasks. Day-to-day, this involves designing and running RL training experiments, evaluating model behavior on agentic reasoning benchmarks, and working closely with my mentor team to refine distillation techniques that make Alexa's underlying language models more capable and efficient reasoners.

 

Describe the process of obtaining your position. When did you hear of the position and submit your application?

I came back to Amazon Alexa AI as a return intern for a second summer. After completing my first internship there in Summer 2025, I stayed in touch with my manager and team, and toward the end of that internship discussed continuing the research into a follow-up summer. I received and accepted the return offer to come back in June 2026 to continue working on self-distillation and reinforcement learning for agentic LLM reasoning.

 

What resources did you use to find your current experience?

LinkedIn or other social media

 

What have you enjoyed the most about your position and organization?

I've most enjoyed getting to work on cutting-edge agentic reinforcement learning problems alongside talented scientists, and seeing research ideas move quickly from concept to real experiments that could impact Alexa's reasoning capabilities.

 

How do you believe you have made an impact through your work?

My work on self-distillation with reinforcement learning aims to make Alexa's underlying LLMs more capable and efficient reasoners on agentic tasks, building on my broader PhD research on decomposition and reinforcement learning for more reliable LLMs.

 

What advice would you give to another student who is seeking an internship or similar experience?

Focus on building a strong research portfolio and don't be afraid to reach out directly to people doing work you admire. Internship and research opportunities often come from persistence, publishing your work, and following up on connections made through faculty and prior collaborators.