Mohsen Tavakoli

I’m a Technical Annotation Specialist, AI Trainer, and AI Tutor focused on helping language models get better—by building the data, rubrics, and evaluation systems that make improvements measurable and reliable. My work is code-centric: I label and evaluate how AI agents plan, write, debug, and complete real software projects.

I’ve worked across the full model-improvement lifecycle, including RLHF/DPO preference ranking, red-teaming, hallucination detection, and reproducible evaluation harnesses for coding and agentic tasks. I design reward signals and verifiers using programmatic checks, LLM-as-judge rubrics, and test-based validation, then generate golden solutions, preference pairs, and targeted edge cases.

I also bring strong end-to-end engineering capability to the same mission. From full-stack automation to production reliability, I’ve built systems for operational visibility, deployed containerized services, and created self-healing workflows that reduce manual intervention.

As an AI Tutor, I translate technical depth into clarity through one-to-one instruction, curriculum design, and prompt engineering—covering programming and ML fundamentals so others can work effectively with AI. I aim to combine rigorous evaluation with practical teaching, so both models and teams improve over time.