Devamanyu Hazarika

AI Research Scientist, Meta Superintelligence Labs (FAIR)

I'm an AI Research Scientist at Meta Superintelligence Labs (FAIR), post-training Muse Spark and enhancing its coding capabilities. My research centers on the post-training and alignment of large language models, and I've spent recent years helping build frontier foundation models. Before Meta I was at Amazon, and I earned my PhD at the National University of Singapore.

// news

Jun 2026 Foundation modelBuilding Muse Spark at Meta Superintelligence Labs, enhancing its coding capabilities.
Jan 2026 AcademicServing as Senior Area Chair for EACL 2026 and ARR October 2025.
Sep 2025 AcademicArea Chair for NeurIPS 2025 and Senior Area Chair for EMNLP 2025.
May 2025 AcademicSenior Area Chair for ACL 2025.
Apr 2025 Foundation modelCore contributor to Llama 4: post-training and coding capabilities.
Dec 2024 Foundation modelFounding member of Amazon Nova 1.0, Amazon's first family of foundation models.

// selected research

all papers →

ICLR (Oral) 2025

Do LLMs Recognize Your Preferences? Evaluating Personalized Preference Following in LLMs

EMNLP (Main) 2025

Split-Merge: Scalable and Memory-Efficient Merging of Expert LLMs

Technical Report 2024

The Amazon Nova Family of Models: Technical Report and Model Card

NAACL 2022

Analyzing Modality Robustness in Multimodal Sentiment Analysis

AAAI 2022

Zero-Shot Controlled Generation with Encoder-Decoder Transformers

ACM Multimedia 2020

MISA: Modality-Invariant and -Specific Representations for Multimodal Sentiment Analysis

ACL 2019

Towards Multimodal Sarcasm Detection (An Obviously Perfect Paper)

ACL 2019

MELD: A Multimodal Multi-Party Dataset for Emotion Recognition in Conversations

AAAI 2019

DialogueRNN: An Attentive RNN for Emotion Detection in Conversations

IEEE Computational Intelligence Magazine 2018

Recent Trends in Deep Learning Based Natural Language Processing

// affiliations

Meta Amazon NUS NTU