Dr. Tasnim Mohiuddin is a Scientist in the Arabic Language Technologies (ALT) group at Qatar Computing Research Institute (QCRI), where he works on pioneering initiatives in the development and enhancement of Large Language Models (LLMs). He is a core contributor to the creation of Fanar, QCRI's flagship LLM. Apart from training LLMs, his current research focuses on domain-specific LLM innovations, including specialized areas such as Code LLMs and Multimodal LLMs.

Before joining QCRI, Dr. Tasnim was a researcher at Huawei Research Center in Singapore, where he concentrated on Multimodal Representation Learning, bridging both linguistic and visual modalities. His research contributions have been consistently recognized through publications in leading academic conferences.

Tasnim completed his doctoral studies at the School of Computer Science and Engineering, Nanyang Technological University (NTU), Singapore, receiving his Ph.D. in June 2022 under the mentorship of Professor Shafiq Joty. His doctoral dissertation was honored with the prestigious Outstanding Ph.D. Thesis Award, reflecting the originality and impact of his research. During his doctoral studies, he also gained valuable experience as a research intern at Meta AI Research, working under the guidance of Professor Philipp Koehn.

1. Agents, Tools, and Planning: Investigating the interplay of intelligent agents, tool-augmented reasoning, and strategic planning to enhance the problem-solving capabilities of language models in complex, real-world tasks.
2. Pre-training and Post-training Strategies for LLMs: Developing innovative methodologies to enhance the efficiency, scalability, and performance of LLMs through advanced pre-training techniques and strategic post-training refinement, aiming to optimize model capabilities, generalizability, and domain-specific adaptability.
3. Multimodal Language Models: Advancing the integration of textual, visual, and auditory modalities to develop sophisticated models capable of holistic, context-aware understanding and generation.
4. Code Language Models: Designing and refining specialized LLMs for programming languages, focusing on generating accurate, efficient, and optimized code to address computational and algorithmic challenges.