Research Scientist – Multimodal Foundation Models

About Mecka AI

Mecka AI is building the data infrastructure layer for robotics and embodied AI.
We design and operate global systems for data capture, data labeling, and hardware-enabled workflows used by leading AI labs and robotics companies to train and validate humanoid and embodied AI systems.
We work closely with frontier robotics teams to bridge real-world data, simulation, learning-based systems, and deployed hardware.

We're looking for an experienced Research Scientist to advance the capabilities of large-scale vision-language and multimodal foundation models.

You'll work on challenging research problems spanning model training, post-training, reinforcement learning, and multimodal reasoning. This is a hands-on role for someone who enjoys experimenting with cutting-edge approaches, improving model performance, and translating research breakthroughs into real-world capabilities.

You'll have the opportunity to work across the model development lifecycle, from data and training strategies to evaluation and optimization, helping push the boundaries of what multimodal AI systems can understand and accomplish.

What You'll Do

• Multimodal Model Research: Research, develop, and improve vision-language models and multimodal foundation-model architectures.

• Post-Training & Reinforcement Learning: Design and implement advanced training strategies, including supervised fine-tuning, preference optimization, and reinforcement learning, to improve model capabilities.

• Model Performance & Optimization: Investigate approaches to improve reasoning, generalization, reliability, and training efficiency.

• Data Quality & Training: Develop strategies for data curation, filtering, and synthetic data generation to improve training outcomes.

• Evaluation & Benchmarking: Design experiments, evaluation frameworks, and benchmarks to measure model performance, identify failure modes, and guide research priorities.

• Large-Scale Experimentation: Work with distributed training infrastructure and large-scale datasets to support research and model development.

• Research Innovation: Explore emerging approaches in multimodal learning, model alignment, and foundation-model training.

What You Bring

Must Have

• Hands-on experience researching, training, fine-tuning, or post-training large-scale foundation models, particularly vision-language or multimodal models.

• Experience working with Qwen, Qwen-VL, or comparable open-weight multimodal foundation models.

• Strong understanding of deep learning architectures, optimization techniques, and modern model training methodologies.

• Practical experience with supervised fine-tuning (SFT), preference optimization, or reinforcement learning approaches such as DPO, GRPO, PPO, or RLHF.

• Strong proficiency in Python and PyTorch, JAX, or comparable deep learning frameworks.

• Experience designing, executing, and analyzing large-scale machine learning experiments.

• Familiarity with distributed training systems and GPU-based model development.

• Strong research instincts, analytical problem-solving skills, and the ability to independently drive complex technical projects.

• Research publications, open-source contributions, or experience within leading AI research organizations.

Nice to Have

• Experience with large-scale model post-training and reinforcement learning for reasoning or alignment.

• Familiarity with multimodal data pipelines involving images, video, or other sensory inputs.

• Experience with reward modeling, synthetic data generation, or advanced model evaluation.

• Experience improving model capabilities through innovations in training strategies, data quality, or model architecture.

A Note on Applying

Studies show women and candidates from underrepresented groups often only apply when they meet 100% of the listed qualifications, while others apply after meeting 60%. If you don't check every box above but believe you can do the job, we encourage you to apply — we're looking for capability and trajectory, not a perfect checklist match.
Inclusive Hiring at Mecka

We are committed to creating an inclusive and supportive candidate experience. Should you require any accommodation whatsoever during the interview process, please inform us without any hesitation. Mecka is dedicated to ensuring equal treatment and opportunity in all phases of recruitment, selection, and employment, in compliance with employment law. We do not discriminate based on gender, race, religion, national origin, ethnicity, disability, gender identity/expression, sexual orientation, veteran or military status, or any other protected category. Mecka is proud to be an equal opportunity employer, fostering a culture of inclusivity and maintaining a work environment that is free from discrimination, harassment, and retaliation.

Use of Artificial Intelligence in Recruitment

Mecka uses artificial intelligence (AI) responsibly to support administrative and efficiency-focused aspects of our recruitment process. This includes activities such as drafting job descriptions, generating interview questions, note-taking and recordings, and supporting sourcing and scheduling workflows. All candidate evaluations, interviews, and hiring decisions are made by members of the Mecka team. While AI tools may assist with screening and assessment, they do not replace human judgment in selection decisions. Our use of AI is intended to streamline routine tasks, improve consistency, and enhance the overall candidate experience. We are committed to upholding principles of fairness, transparency, and accountability in all hiring activities. Mecka regularly reviews its recruitment practices to mitigate bias and to ensure alignment with applicable laws and evolving best practices.