LLM Research Intern

About the Role

Sony Research India is seeking a highly motivated individual to join their research team. This role involves deep research and rapid prototyping with Large Language Models (LLMs), with a focus on multilingual understanding, translation, reasoning, and domain adaptation. The internship aims to investigate how modern LLMs can be adapted, analyzed, and improved for multilingual and entertainment-domain applications, with the potential to generate publications, prototypes, and production impact. This is a research-intensive role for individuals interested in understanding model behavior and developing new methods.

Key Responsibilities

• Conduct research on LLMs for multilingual language understanding and generation, cross-lingual transfer, language interference, model behavior, and multilingual reasoning, particularly across Indian and international languages.
• Explore domain adaptation for entertainment content, including subtitles, dialogues, scripts, and conversational language.
• Explore reasoning and adaptive computation for translation, including when and how additional reasoning or context should be used.
• Explore reinforcement learning, preference optimization, bandits, or other learning-based approaches for improving LLM/NMT behavior.
• Rapidly prototype research ideas using modern LLM/NMT frameworks and evaluate them on research benchmarks and real-world entertainment data.
• Collaborate with research scientists to formulate hypotheses, implement experiments, analyze results, and contribute to research papers, patents, and internal prototypes.
• Fine-tuning LLMs (e.g., LoRA, QLoRA, Instruction Tuning).

Work Location

• Remote within India.

Duration

• This paid internship will be for six months, 9:00 to 18:00 (Monday to Friday), starting November first week of 2026.

Requirements & Tech Stack

• Python
• PyTorch
• LLMs
• NMT
• LoRA
• QLoRA
• PEFT
• Instruction Tuning
• Deep Learning
• Reinforcement Learning
• Preference Optimization
• Bandits

Qualifications

• Ph.D. or Masters in Computer Science background, or Bachelor's with hands-on experience with Python in fine-tuning LLMs for Language Processing applications.
• Excellent coding skills, especially in Python and PyTorch.
• Hands-on experience with fine-tuning LLMs (e.g., LoRA/QLoRA, PEFT, Instruction Tuning).
• Prior self-driven project experience in Deep Learning.
• In-depth knowledge of the state-of-the-art in LLMs and foundation models.

Preferred Skills

• Strong mathematical aptitude.
• Publications in top-tier conferences like ICML, NeurIPS, ICLR, AAAI, and ACL.