LLM Research Intern

About this role

About the opportunity: - Join scientists researching multilingual LLM behaviour, translation, reasoning and entertainment-domain adaptation. - Sony's global research develops technology supporting electronics, entertainment and finance. Research activities: - Investigate Indian and international languages, cross-language transfer, language interference, understanding and generation. - Adapt models for subtitles, dialogue, scripts and conversational material. - Study contextual reasoning and adaptive computation in translation. - Explore reinforcement learning, preference optimisation, bandits and related approaches when useful. - Prototype with LLM/NMT frameworks and evaluate against benchmarks and entertainment datasets. - Form hypotheses, implement experiments, interpret results and contribute papers, patents and internal prototypes. - Apply LoRA, QLoRA and instruction tuning. Required background: - Have PhD/master's-level computing education, or a bachelor's qualification with practical Python-based LLM language-processing experience. - Bring strong Python/PyTorch programming, PEFT and fine-tuning experience, independent deep-learning projects and current foundation-model knowledge. - Mathematical strength and publications at ICML, NeurIPS, ICLR, AAAI or ACL are preferred. Practical details: - The focus is developing research methods and understanding models, rather than primarily integrating existing models. - Work remotely within India for six paid months, starting in November 2026's first week. - Hours are 9:00–18:00, Monday–Friday; payment amount is undisclosed. - Curiosity, diversity, integrity, sustainability and equal opportunity guide the organisation.