Hey, I’m Mohammed. I’m an AI/ML Research Engineer based in London, with a BSc in Computer Science with Mathematics from University College Dublin.

The work that pulls me in is where there are major gaps between what exists and what’s needed: RL environments, agent observability, multimodal AI, and domain specific systems. I like finding problems that have not been solved well yet and building real solutions for them.

Experience

My work has covered RL environments, agent observability, RLVR evaluation, and RL tuned conversational behavior. That includes post training loops with automated evals, multimodal workflows, and fine tuned domain specific models.

I co-founded Taqriry.ai, where I led ML and product engineering for an AI notetaker. The work included RAG, summarization, multimodal workflow, and inference systems for production and on prem deployment.

At IBM, I worked on Watsonx multimodal AI systems and agent pipelines for enterprise workflows, including HR automation and an LLM legal chatbot.

Research Focus

I’m interested in pushing AI beyond fixed context windows, designing RL environments for agents, and using post-training to make models outperform closed-source systems. The focus is taking foundation models to production, building scalable systems, and using strong evaluation and observability loops to improve them. A lot of my work sits around RL for language and reasoning (RLHF, RLVR, PPO, GRPO), multimodal systems, and agent infrastructure where training shapes real problem-solving ability. More recently, this has expanded toward the physics side: humanoid robotics and autonomous systems.