TII — Technology Innovation Institute — a publicly funded research institute in Abu Dhabi, UAE, home to leading scientists, engineers, mathematicians, and researchers from across the globe — is seeking a VLM Engineer for its Artificial Intelligence and Digital Research Centre. Joining the Extreme-Scale Language Model team — the same team that developed the globally recognised Falcon models — this researcher-engineer role conducts vision model and data ablation studies, develops robust model evaluation protocols, and engages in model training that integrates LLMs with vision models like CLIP, using PyTorch and distributed training techniques. A PhD in Deep Learning is mandatory, with a publication record in top-tier conferences highly desirable.
About TII — Technology Innovation Institute Abu Dhabi UAE
Institute: TII — Technology Innovation Institute — a publicly funded research institute in Abu Dhabi transforming complex research challenges into pioneering technology prototypes that advance society through rigorous scientific discovery
AI & Digital Research Centre: The Extreme-Scale Language Model team has developed the globally recognised Falcon models and is continuing into cutting-edge applied research in large language models, vision-language alignment, and multimodal AI
Research Scope: AI, advanced materials, autonomous robotics, cryptography, digital security, directed energy, quantum computing, and secure systems — among the most ambitious and comprehensive applied research mandates of any institution in the Middle East
Team Context: A diverse community of leading scientists, engineers, mathematicians, and researchers from across the globe — state-of-the-art facilities and collaboration with leading international research institutions
Why This VLM Engineer Role at TII Stands Out
Falcon Team Continuation: Join the team that built the Falcon models — one of the most widely downloaded and benchmarked open-source LLM families in the world — and contribute to the next generation of vision-language model research at TII’s Extreme-Scale Language Model team
Vision-Language Frontier Research: Conduct comprehensive vision and data ablation studies, develop evaluation protocols, and integrate LLMs with vision models like CLIP — working on the research problems at the intersection of NLP and computer vision that define the direction of multimodal AI
Abu Dhabi World-Class Research Environment: Access to state-of-the-art computational facilities, a globally diverse research community, and collaboration with leading international institutions — in a publicly funded research institute with the resources to pursue genuinely ambitious AI research programmes
Global Research Impact: Your work contributes to research publications, model releases, and applied AI innovations that reach the global research and technology community — at an institute with the profile, resources, and team calibre to produce world-leading results
Position Overview — VLM Engineer, Extreme-Scale Language Model Team
This VLM Engineer at TII’s Artificial Intelligence and Digital Research Centre in Abu Dhabi conducts comprehensive ablation studies on vision models to assess the impact of various components and configurations, collaborates with researchers to analyse and report on the effectiveness of different model architectures and settings, partners with team members to perform data ablation studies identifying optimal data types and structures for training vision-language models, analyses the impact of different data inputs on model performance with particular focus on vision-language alignment, develops and implements robust evaluation protocols for vision-language models, assesses model performance across diverse benchmarks and real-world scenarios, and engages in model training with emphasis on integrating LLMs with vision models like CLIP — in the team that produced the globally recognised Falcon model family.
Why This Role Matters: As VLM Engineer at TII’s Extreme-Scale Language Model team in Abu Dhabi, you are contributing to one of the most consequential research programmes in multimodal AI currently being pursued anywhere in the world outside the largest technology company research laboratories. When your vision model ablation study correctly identifies that a specific architectural component is contributing disproportionately to cross-modal alignment quality — and that finding changes the team’s model design decisions for the next training run at billion-parameter scale, your data ablation research correctly identifies that a particular class of image-text training data is degrading rather than improving vision-language grounding in ways that were not visible in the aggregate benchmark scores, or your evaluation protocol correctly surfaces a benchmark gap between headline performance and real-world vision-language reasoning capability that the existing evaluation suite had systematically missed — you are not running experiments. You are generating the empirical insights that determine the design of the next generation of vision-language models that the global AI research community will build upon. At TII, research at this scale and quality has already produced the Falcon models. Your contribution to the VLM programme determines what comes next.
Key Responsibilities
Vision Model Ablation Studies & Architecture Analysis
- Conduct comprehensive ablation studies on vision models — systematically varying and isolating individual components, architectural choices, and configuration parameters to assess their specific contribution to vision-language model performance, alignment quality, and generalisation behaviour across benchmarks and real-world evaluation scenarios
- Collaborate with researchers to analyse and report on the effectiveness of different model architectures and settings — contributing ablation study findings to the team’s understanding of what design choices matter most in vision-language model development at extreme scale
- Engage in model training with emphasis on integrating LLMs with vision models like CLIP — contributing to the engineering and research challenges of multimodal model construction where language and vision representations must be aligned, trained, and evaluated in a coherent framework that produces genuine cross-modal understanding
Data Ablation Research & Vision-Language Alignment
- Partner with team members to perform data ablation studies identifying optimal data types, data structures, and data mixture strategies for training vision-language models — addressing one of the most empirically challenging and commercially consequential research questions in multimodal AI: what data, and how much of it, actually makes a VLM better
- Analyse the impact of different data inputs on model performance — with particular focus on vision-language alignment quality, investigating how different image-text training data sources, caption quality levels, and data curation approaches affect the resulting model’s ability to ground language in visual perception
- Apply expertise in dataset curation and processing for vision and language tasks — handling the large-scale, heterogeneous multimodal datasets that extreme-scale VLM training requires, with the data engineering rigour needed to ensure training data quality translates into model capability improvements
Model Evaluation Protocol Development & Benchmarking
- Develop and implement robust evaluation protocols for vision-language models — designing evaluation frameworks that go beyond existing benchmark suites to assess model capabilities across the diverse, realistic, and challenging scenarios that reveal genuine vision-language understanding rather than benchmark-optimised surface performance
- Assess model performance across diverse benchmarks and real-world scenarios — applying systematic, rigorous evaluation methodology that provides the team with accurate, comparable, and interpretable performance characterisations to guide iterative model development decisions
- Apply strong analytical skills in conducting ablation studies and evaluating model performance — bringing the quantitative analytical rigour and experimental design discipline that distinguishes high-quality machine learning research from exploratory experimentation
Qualifications, Experience & Research Profile
Essential Requirements
- PhD degree in Deep Learning, Machine Learning, Computer Vision, Natural Language Processing, or a closely related field — this is a mandatory qualification; candidates without a completed PhD in a relevant deep learning research area will not be considered for this role
- Proven track record of research and development in vision-language models — demonstrating genuine empirical and theoretical contribution to the VLM research field, not merely application of existing methods
- Expertise in machine learning — particularly in vision-language models and LLMs — with deep understanding of model architectures like CLIP and their application in vision-language tasks
- Proficiency in distributed training techniques and multi-GPU optimisation — essential for training vision-language models at the extreme scale that TII’s research programme targets
- Experience with deep learning frameworks — particularly PyTorch — applied to research-grade model development and experimentation at scale
- Strong analytical skills for conducting ablation studies and evaluating model performance — including experimental design, statistical analysis of model behaviour, and systematic evaluation across diverse benchmarks
- Familiarity with dataset curation and processing for vision and language tasks — handling large-scale multimodal training datasets with the data engineering rigour that extreme-scale VLM training requires
Highly Desirable
- Publication record in top-tier AI and ML conferences — NeurIPS, ICML, ICLR, CVPR, ECCV, ACL, EMNLP, or equivalent venues — demonstrating the research quality and communication standard that TII’s research programme upholds
About TII — Where Pioneering Research Meets Global Impact
Technology Innovation Institute (TII) is a publicly funded research institute based in Abu Dhabi, UAE — home to a diverse global community of leading scientists, engineers, mathematicians, and researchers who transform problems into pioneering research and technology prototypes that advance society. TII’s Artificial Intelligence and Digital Research Centre is home to the Extreme-Scale Language Model team — which developed the Falcon models, among the most widely downloaded and benchmarked open-source LLMs in the world — and is now pursuing the next generation of vision-language model research and applied AI innovation. With state-of-the-art facilities, collaboration with leading international research institutions, and the resources of a publicly funded institute committed to world-leading research across AI, quantum computing, cryptography, autonomous robotics, and beyond, TII offers a research environment that is genuinely exceptional for the most ambitious AI researchers in the world. The VLM Engineer role is your opportunity to contribute to that programme at the frontier of multimodal AI.
Career Excellence: Conduct VLM ablation studies, develop evaluation protocols, and integrate LLMs with vision models — VLM Engineer at TII, Abu Dhabi UAE, Falcon team 2026.
Who Should Apply?
- PhD Deep Learning Researchers — Vision-Language Models: With a completed PhD and research track record in vision-language model development — ablation studies, VLM training, evaluation protocol design, and cross-modal alignment research at the frontier of multimodal AI
- CLIP & LLM Integration Specialists — PyTorch: With hands-on experience training and evaluating vision-language model architectures including CLIP and LLM-vision integration — using distributed PyTorch at multi-GPU scale in a research or applied AI engineering context
- Multimodal AI Researchers — NeurIPS/CVPR/ICLR Publications: With a publication record in top-tier AI and ML venues — bringing research communication standards and empirical rigour consistent with TII’s position as a globally recognised AI research institute
- Dataset Curation & Ablation Study Specialists — VLM Training: With experience designing and executing vision-language data ablation studies and large-scale dataset curation workflows — contributing to the empirical research agenda of an extreme-scale language model team
- Abu Dhabi-Based AI Research Engineers — Global Research Ambition: Seeking a VLM Engineer role at TII — one of the world’s most resourced and ambitious publicly funded AI research institutes — where world-class colleagues, state-of-the-art infrastructure, and the legacy of the Falcon model programme provide the environment to do research that genuinely advances the frontier of vision-language AI
Recently Opening job👇
