Shape vision-language intelligence at Mecka AI by improving multimodal reasoning, model performance, and training efficiency.

As an experienced Research Scientist, you will tackle research problems spanning model training, post-training, reinforcement learning, and multimodal reasoning. Your work will translate emerging approaches into stronger foundation models. You will also guide experimentation, data strategies, evaluation, and optimization across the model lifecycle.

Key Responsibilities:

- Develop vision-language and multimodal foundation-model architectures
- Explore multimodal learning, alignment, and foundation-model training
- Improve model reasoning, generalization, reliability, and efficiency
- Develop data curation, filtering, and synthetic-data strategies




- Design experiments and frameworks measuring model performance

Requirements:

- Research experience with vision-language or multimodal foundation models
- Practical SFT, DPO, GRPO, PPO, or RLHF experience
- Strong knowledge of modern model training methodologies
- Familiarity with distributed training and GPU-based development
- Strong analytical skills and independent technical project leadership

Apply deep learning research and multimodal reasoning expertise to advance reliable, productive foundation models at Mecka AI.

📌 Vision Language Research Scientist (Richmond Hill)
🏢 mecka
📍 Richmond Hill